Validity of Scandura and Ragins (1993) Multidimensional Mentorin

Embed Size (px)

DESCRIPTION

Validity of Scandura and Ragins (1993) Multidimensional Mentorin

Citation preview

  • University of MiamiScholarly Repository

    Management Faculty Articles and Papers Management

    1-1-2004

    Validity of Scandura and Ragins' (1993)Multidimensional Mentoring Measure: AnEvaluation and RefinementStephanie L. CastroFlorida Atlantic University

    Terri A. Scandura PhDUniversity of Miami, [email protected]

    Ethlyn A. WilliamsFlorida Atlantic University

    This Conference Proceeding is brought to you for free and open access by the Management at Scholarly Repository. It has been accepted for inclusionin Management Faculty Articles and Papers by an authorized administrator of Scholarly Repository. For more information, please [email protected].

    Recommended CitationCastro, Stephanie L.; Scandura, Terri A. PhD; and Williams, Ethlyn A., "Validity of Scandura and Ragins' (1993) MultidimensionalMentoring Measure: An Evaluation and Refinement" (2004). Management Faculty Articles and Papers. Paper 7.http://scholarlyrepository.miami.edu/management_articles/7

  • Multidimensional Mentoring Measure 1

    Validity of Scandura and Ragins (1993) Multidimensional Mentoring Measure:

    An Evaluation and Refinement

    Stephanie L. Castro

    Department of Management

    College of Business

    Florida Atlantic University

    2912 College Avenue

    Davie, FL 33314

    (954)236-1350

    (954) 236-1295 fax

    [email protected]

    Terri A. Scandura

    Department of Management

    School of Business

    University of Miami

    414 Jenkins Building

    Coral Gables, FL 33124-9145

    (305) 284-3746

    (305) 284-3655 fax

    [email protected]

    Ethlyn A. Williams

    Department of Management

    College of Business

    Florida Atlantic University

    777 Glades Road

    Boca Raton, FL 33431

    (561) 297-2357

    (561) 297-2675 fax

    [email protected]

    A previous version of this paper was presented at the Southern Management Association

    Meetings, San Antonio, Texas, 2004.

    We would like to thank Chet Schriesheim for his constructive comments on an earlier draft.

  • Multidimensional Mentoring Measure 2

    Validity of Scandura and Ragins (1993) Multidimensional Mentoring Measure:

    An Evaluation and Refinement

    Abstract

    The establishment of a mentoring relationship can be important to an individuals career for

    multiple reasons. However, in order to study this construct, we must be able to accurately

    measure it. In this paper, three separate studies were conducted to examine and refine Scandura

    and Ragins (1993) multidimensional mentoring measure. In Study 1, an empirical assessment of

    the content validity of the measure was conducted. The convergent and discriminant validity,

    reliability, and item-total correlations were then examined in Study 2, and the measure was

    reduced to nine items. The convergent and discriminant validity, reliability, and item-total

    correlations of this reduced measure were then assessed. In Study 3, these same evaluations were

    repeated. Concurrent validity was also evaluated by relating the measure to organizational

    outcomes. Based on these studies, initial indications of the validity of the revised 9-item

    measure are good. Strengths, limitations, and directions for future research are discussed.

    (149 words)

    Keywords: Measurement design, reliability and validity, survey research, construct validity,

    mentoring

  • Multidimensional Mentoring Measure 3

    Mentoring relationships have been discussed in the academic and practitioner

    communities as critical for the development of protgs, as these relationships generally have a

    positive impact on individual and organizational outcomes. A recent meta-analysis demonstrated

    that protgs experience substantial career benefits, including positive objective outcomes such

    as increased compensation and promotions as well as subjective outcomes such as increased job

    satisfaction and reduced turnover intentions (Allen, Eby, Poteet, Lentz, & Lima, 2004).

    However, the field is still in its relative infancy, as evidenced by the small number of studies

    included in the meta analysis (i.e., 43). As we move forward and attempt to improve our

    understanding of the nature of the mentoring relationship (Underhill, 2006), we need to be

    concerned with the validity of the instruments we use to measure the mentoring construct.

    The importance of obtaining evidence supporting the construct validity of measures used

    in empirical research has been emphasized by many (e.g., Nunnally & Bernstein, 1994; Schwab,

    1980). Unfortunately, researchers all too often do not heed this advice, and use measures in

    organizational studies that lack sufficient evidence of construct validity (Cook, Hepworth, Wall,

    & Warr, 1981; Podsakoff & Dalton, 1987; Schriesheim, Powers, Scandura, Gardiner, & Lankau,

    1993; Stone-Romero, 1994). This violates the basic principles of scientific research, for without

    accurate instruments to measure constructs, we can have little confidence in the results. As

    Korman (1974, p. 194) pointed out, The point is not that accurate measurement is nice. It is

    necessary, crucial, etc. Without it we have nothing.

    In this study we evaluate the validity of Scandura and Ragins (1993) multidimensional

    mentoring measure. The measure has 15 items, and is comprised of three scales measuring three

    dimensions: career support, psychosocial support, and role modeling. Although this measure has

    been used in a number of studies, there is limited evidence of its validity. The majority of the

  • Multidimensional Mentoring Measure 4

    validity-related evidence to date addresses concurrent validity. Thus far, Scandura and Ragins

    mentoring measure and/or its subdimensions have been positively related to personal learning

    and job satisfaction (Lankau & Scandura, 2002), value similarity and supportiveness (Neilson,

    Carlson, & Lankau, 2001), promotion rate and salary level (Scandura, 1992), and feminity

    (Scandura & Ragins, 1993). The measure and/or its subdimensions have been negatively

    associated with role ambiguity and intent to leave (Lankau & Scandura, 2002), family

    interference with work (Nielson et al., 2001), and organizational rank and mentorship experience

    (Scandura & Ragins, 1993).

    Minimal support for the factor structure exists. Scandura and Ragins (1993) found

    support for the three factor structure in their initial study. Williams (1999) also found support for

    three factors. Arguably, given the limited evidence thus far, more information regarding the

    validity of Scandura and Ragins (1993) measure is needed.

    Thus, the primary purpose of this paper is to evaluate and refine Scandura and Ragins

    (1993) mentoring measure through a comprehensive, programmatic approach. Quantitative,

    empirical analyses of content validity, convergent validity, and discriminant validity are

    conducted, and reliability and item-total correlations are examined. Based on our results, a 9-

    item version of Scandura and Ragins (1993) measure is proposed, and concurrent validity is then

    assessed. The 9-item measure is then evaluated in a separate confirmatory sample.

    Study 1

    Study 1 was conducted to empirically assess the content validity of Scandura and Ragins

    (1993) mentoring measure. The quantitative methodology described by Schriesheim et al. (1993)

    was employed. The mentoring items and the three corresponding subdimension definitions were

    presented to judges, as well as a fourth category, defined as Behaviors which clearly do not fall

  • Multidimensional Mentoring Measure 5

    into the other categories. The judges were asked to evaluate the items consonance with the

    theoretical definitions.

    Method

    Sample

    Students attending a medium-sized private university in the southeastern United States

    were used as judges. The use of students for this type of analysis should not be problematic,

    since they are not asked about actual work or mentoring experience; they are only asked to make

    semantic judgments about item fidelity to definitions (cf. Schriesheim et al., 1993). The total

    sample size was 169, and 50% were male. The average age was 24 years old and 96% were

    employed or had been previously employed. The sample was 40% Caucasian, 31% Hispanic,

    10% African American, 6% Asian, and 12% other.

    Measure

    Scandura and Ragins (1993) measure, comprised of three scales measuring career

    support, psychosocial support, and role modeling, was employed. Judges were given the 15

    items, the three subdimension definitions, and the none of the above category. (A copy of the

    form is available from the authors upon request.)

    Analyses

    The data from the content adequacy assessment was analyzed using principal axis factor

    analysis with oblique rotations (delta = 0). The data were analyzed in two ways: (1) factors with

    an eigenvalue greater than or equal to 1 were extracted, and (2) the number of factors was

    specified (based on theory). For this latter procedure, three factors were specified (one each for

    career support, psychosocial support, and role modeling).

  • Multidimensional Mentoring Measure 6

    Results

    The results for the two analyses (extracting factors with eigenvalues greater than or equal

    to 1.0 or specifying the number of factors) were identical. The first four unrotated eigenvalues

    for the Scandura and Ragins (1993) 15-item measure were 3.45, 2.55, 1.89, and .86, supporting

    the extraction of three factors. The total variance explained was 41%. The factor loadings

    (pattern coefficients) are presented in Table 1. Following the recommendation of Ford,

    MacCallum, and Tait (1986), loadings of .40 or greater on the intended theoretical factor were

    considered meaningful for interpretation, and cross-loadings of .30 or greater were considered

    excessive. All items loaded .40 or stronger on the intended factor, and there were no cross-

    loadings of .30 or greater.

    --------------------------------------

    Insert Table 1 about here

    --------------------------------------

    Discussion

    The results of the factor analysis strongly support the content validity of Scandura and

    Ragins (1993) measure. All of the items exceeded the criteria set forth by Ford et al. (1986).

    Importantly, there were no cross loadings, indicating that the items are unidimensional.

    Additionally, the fourth category, None of the above, was not supported. Thus, the items can

    be considered to strongly reflect their intended theoretical constructs.

    Study 2

    The purpose Study 2 was to assess scale reliability, concurrent validity, and convergent

    and discriminant validity. Concurrent validity was evaluated by relating the mentoring measure

    and subscales to job satisfaction, organizational commitment, and anxiety. The two former

    constructs were chosen as a significant positive correlation with mentoring is expected, based on

  • Multidimensional Mentoring Measure 7

    theory and previous empirical analyses (e.g., Allen et al., 2004). General anxiety was also

    measured, but was expected to have only a weak (negative) relationship with mentoring.

    Convergent validity was evaluated by examining (1) the ability of the hypothesized factor

    structure of each scale to account for a large amount of covariation among the measures items,

    (2) the statistical significance and size of the item loadings on the hypothesized factors, (3) the

    proportion of variance in the latent variables accounted for by the indicators using Fornell and

    Larckers (1981) average variance indicator (vc()), and (4) the correlations with Ragins and

    McFarlins (1990) mentoring measure. Discriminant validity was assessed by (1) testing that

    each of the correlations among the scales was significantly less than 1, and (2) comparing vc()

    with scale intercorrelations. Fornell and Larcker (1981) argued that to demonstrate discriminant

    validity, the average variance in a latent constructs indicators should be greater than the square

    of the constructs correlations with other latent constructs. Due to the somewhat disappointing

    outcomes of the Study 2 analyses, the scale was reduced to a 9-item measure, and the analyses

    repeated.

    Method

    Sample

    Employed MBA students attending classes at a medium-sized private university in the

    southeastern United States were surveyed. The total sample size was 474; of these, 255 (54%)

    indicated having had a mentor at some time in their careers. Only those who had been mentored

    were used in the analyses. Listwise deletion of missing data resulted in a sample of N = 246.

    This sample was used in the Lisrel analyses, evaluating convergent and discriminant validity, and

    the reliability and item-total analyses. However, for the analyses including the Ragins and

    McFarlin (1990) measure, missing data resulted in a sample of N = 160. This smaller sample

  • Multidimensional Mentoring Measure 8

    was only used in the regression analyses comparing the two measures (reported in Table 8) and

    the correlations (presented in Table 9).

    In the mentored sample, 54.7% were male, and the average age was 28.8 years old. The

    sample was 52% Caucasian, 7.4% African American, 23% Hispanic, 6.3% Asian, and 3.1%

    indicated other. On average, the respondents had 4.8 years of experience prior to joining their

    current organization, and 3.3 years experience in their current position. Respondents jobs

    ranged from first line supervisor to CEO.

    Measures

    Scandura and Ragins (1993) measure was used. The coefficient alpha reliability

    estimate for this administration of the original 15-item measure was .93, while the reliability of

    the scores on the revised 9-item measure was .91 and the reliability of the scores on the 3-item

    revised scales ranged from .82 to .85.

    The 33-item measure developed by Ragins and McFarlin (1990) was used as an

    alternative measure of the mentoring construct. The measure is comprised of eleven scales, each

    with three items: sponsorship, coaching, protection, challenge, exposure, friendship, social,

    parent, role modeling, counseling, and acceptance-and-confirmation. The social and parent

    scales were developed for specific mentoring situations (cross-gender), and are not directly

    comparable to the Scandura and Ragins (1993) scale. Thus, the six items measuring these two

    dimensions were eliminated from the analyses in this study. 1 The coefficient alpha for the data

    collected on the 27-item scale was .91.

    Job satisfaction was measured using the 20-item short form of the Minnesota Satisfaction

    Questionnaire (MSQ; Weiss, Dawis, England, & Lofquist, 1967). The MSQ was employed due

    to its careful development and refinement, as well as the considerable evidence which supports

  • Multidimensional Mentoring Measure 9

    its validity and reliability (e.g., Wanous, 1974; Weiss et al., 1967). The coefficient alpha

    reliability estimate in this sample was .86.

    The Organizational Commitment Questionnaire (OCQ; Porter, Steers, Mowday, &

    Boulian, 1974) was used to measure organizational commitment. There have been numerous

    studies which have investigated the construct validity of the OCQ, and while there is evidence

    supporting its validity, there are also concerns. In particular, some have questioned the factor

    structure of the measure and recommended a 9-item form (e.g., Mathieu & Zajac, 1990).

    However, the 9-item version of the OCQ only measures affective commitment. Since our desire

    was to capture the entire theoretical construct of organizational commitment, we used the

    complete OCQ in these analyses. Coefficient alpha in this sample was .85.

    Anxiety was measured by the State-Trait Anxiety Inventory (Speilberger, Gorush &

    Lushene, 1970). Respondents were asked to indicate how they generally felt. Sample items

    include I feel nervous, and I am worried. This measure has been used in over 150 studies,

    and substantial evidence supports its validity (e.g., Rule & Traver, 1983; Sherwood &

    Westerback, 1983). The coefficient alpha reliability estimate in this data was .90.

    Analyses

    Convergent and discriminant validity. The convergent and discriminant validity of the

    measure was evaluated using Lisrel 8.51 (Jreskog & Srbom, 1993). Sample covariances

    served as input for the maximum likelihood estimations, each item was specified as loading on

    only one factor, the errors were specified as uncorrelated among themselves, and the latent trait

    factor correlations were freely estimated.

    Four rival models were tested. A single factor model with all the items loading on one

    global mentoring factor was tested. Two alternate two factor models were tested. In the first two

  • Multidimensional Mentoring Measure 10

    factor model (Model 2a), the role modeling and career support items loaded on one factor while

    the psychosocial support items loaded on a second factor. In the second two factor model

    (Model 2b), the role modeling and psychosocial support items loaded on one factor while the

    career support items loaded on a second.2 Finally, a three factor model was tested, with the

    career, psychosocial, and role modeling items each loading on their respective factors.

    Scale reliabilities and item-total correlations. Coefficient alpha internal consistency

    scale reliabilities and item-total correlations were examined for each measure and its scales.

    Reduced 9-item measure. Based on the item analyses, scale reliabilities, content validity

    results, and factor analysis results, Scandura and Ragins (1993) measure was reduced to nine

    items, three items per scale. Scale reliabilities, item-total correlations, and convergent and

    discriminant validity were then assessed on the reduced measure and scales.

    Comparison of the two mentoring measures. An additional analysis used to support the

    construct validity of the reduced 9-item mentoring measure was an investigation of the ability of

    the Ragins and McFarlin (1990) measure to add to explained variance in outcome measures over

    the 9-item measure. Since both measures assess the same global construct (mentoring), neither

    should add explanatory power over the other. As Liden and Maslyn (1998) did comparing the

    LMX-7 scale (Scandura & Graen, 1984) with their newly-constructed LMX-MDM scale, we ran

    a series of hierarchical regression analyses. Using job satisfaction, organizational commitment,

    and anxiety as dependent variables (each in separate regression analyses), the revised 9-item

    measure was entered first. In the second step, the 27-item Ragins and McFarlin (1990) scale was

    added. The change in R2 was then evaluated to determine whether the Ragins and McFarlin

    (1990) measure was able to explain any additional variance in the outcome variables. The

    analyses were then repeated, reversing the order of entry.

  • Multidimensional Mentoring Measure 11

    Concurrent validity. Concurrent validity was examined by looking at correlations

    between the mentoring measure and its three scales and outcome variables (job satisfaction,

    organizational commitment, and anxiety).

    Results

    Convergent and discriminant validity

    The fit statistics for the four alternative models are presented in Table 2. The

    hypothesized three factor model fit best, and significantly better than either of the two factor

    models. The chi-square difference between the three factor model and the next-best fitting model

    (model 2a) was 66.16 (df = 2), significant at the p .001 level. However, the chi-square statistic

    was statistically significant (2 = 297.82, df = 87) and the ratio of the chi-square to degrees of

    freedom approached 3.5, indicating only a moderate fit (Bentler & Bonett, 1980). The fit indices

    also indicated a moderate fit, with the SRMSR = .067, the GFI = .84, the TLI = .88, the NFI =

    .86, and the CFI = .90. While the fit indices for the three factor model were better than the

    alternative models, only the CFI met the .90 standard (Medsker, Williams, & Holahan, 1994).

    The root mean square error of approximation (RMSEA) for the three factor model was .11,

    indicating a poor fit (Browne & Cudeck, 1993). Additionally, model comparisons were made

    using the 90% confidence interval of the RMSEA. Given that the confidence interval for the

    three factor model overlaps with the confidence interval for model 2a, we cannot conclude that

    the three factor model has a significantly better fit (Browne & Cudeck, 1993).

    --------------------------------------

    Insert Table 2 about here

    --------------------------------------

    The completely standardized factor loadings and error variances for the 15-item measure

    are presented in Table 3. All of the loadings were statistically significant (p .05), with an

  • Multidimensional Mentoring Measure 12

    average factor loading of .74. However, there were several factor loadings that were relatively

    low. The loadings for item 3 (.69), item 11 (.63), and 14 (.58) were all below .70.

    The disattenuated correlations and average variances are presented in Table 4. The

    disattenuated correlations ranged from .74 to .83, with an average of .77. Although the

    correlations were high, they were significantly less than 1.0 (p .01). Additionally, they were

    not high enough to suggest a second order factor existed. Finally, in looking at the average

    variance (vc()), the hypothesized dimensions (latent constructs) accounted for a large proportion

    of variance (on average) in their indicators. The average variance accounted for ranged from .52

    to .58, with an average of .55. In comparing vc() to the disattenuated correlations, three of the

    six comparisons did not fit the criteria. The vc() for the career support scale was .52, which was

    less than the square of the disattenuated correlation between career support and psychosocial

    support (.55), and less than the square of the disattenuated correlation between career support and

    role modeling (.69). Also, the vc() for the role modeling scale was .56, which was less than the

    square of the disattenuated correlation between career support and role modeling (.69).

    -------------------------------------------

    Insert Tables 3 and 4 about here

    -------------------------------------------

    Finally, the correlation between the 15-item measure and Ragins and Mc Farlins (1990)

    27-item measure (excluding the social and parent scales) was r = .73 (p .01; see Table 9).

    Based on these results, there appeared to be some evidence of the convergent and

    discriminant validity of Scandura and Ragins (1993) 15-item measure. However, the moderate

    overall model fit, the poor model comparison results, and the average variance comparisons

    indicated that the measure could be improved.

  • Multidimensional Mentoring Measure 13

    Scale Reliabilities and Item-total Correlations

    The coefficient alpha internal consistency reliability estimate for the 15-item measure was

    .93. Item-total correlations were moderately strong, generally ranging from .60 to .78, with two

    items below .60 (for item 11, rit = .55, and for item 14, rit = .49). The reliability estimates for the

    data collected for the career support, psychosocial support, and role modeling scales were .88,

    .87, and .83, respectively. Item analyses indicated that deleting single items would not increase

    the scale reliabilities for either the full measure or the three scales.

    Revised 9-item Measure

    Based on the results of the analyses, we felt the 15-item Scandura and Ragins (1993)

    measure needed to be modified. The measure was reduced to nine items, hereinafter referred to

    as the Mentoring Functions Questionnaire, or MFQ-9. Three items were retained for each

    dimension (career support, psychosocial support, and role modelingthe retained items are

    listed in Table 6). The theoretical construct definitions were taken into consideration in

    determining which items to retain, as were the item-total correlations, factor loadings, and

    content adequacy results. The rationale behind the retention/elimination of particular items is

    discussed below.

    For the career support scale, rather than retain items that focused on one of the particular

    subdimensions (e.g., coaching), we chose to retain items that were more general in nature. For

    example, we dropped the item My mentor advised me of promotional opportunities because

    the situation may preclude the availability of such opportunities (i.e., there may be no

    promotional opportunities in the organization). The lack of promotional opportunities, however,

    does not prevent a mentor from providing career support in other ways. The other items

    eliminated were similar, in that the mentor may or may not have had the opportunity to engage in

  • Multidimensional Mentoring Measure 14

    the behavior due to circumstances. Notably, the item with the low loading from the confirmatory

    factor analysis (item 3) was specific (rather than general), and thus it was eliminated.

    For the psychosocial support scale, we eliminated items that were possibly biased in that

    they could only apply to mentor-protg relationships occurring at the same physical location.

    The items I often go to lunch with my mentor and I socialize with my mentor after work

    were dropped, as it may be physically impossible to engage in these behaviors (different physical

    location, different working hours, inability to take off at the same time, etc.), yet psychosocial

    support could still occur. Additionally, the former item obtained a low confirmatory factor

    loading, further contributing to the reasoning for elimination.

    Finally, for the role modeling scale, we eliminated the one item that was not behavioral.

    Item 14 (which also had a low loading in the confirmatory factor analysis), I respect my

    mentors knowledge of the profession, was not related to any behavior on the part of the mentor,

    and thus was not something the protg could model. The three items retained referred to

    things that could be imitated (i.e., motivating others, teaching others), or to the actual act of

    trying to imitate the mentors behavior.

    The analyses (confirmatory factor analyses, reliabilities, and item-total correlations) were

    then repeated on the revised 9-item measure, or the MFQ-9.

    Convergent and discriminant validity. The four alternative models were tested using the

    MFQ-9, and the fit statistics are presented in Table 5. The hypothesized three factor model fit

    best. The chi-square statistic was statistically significant (2 = 79.3, df = 24, p .001), and the

    ratio of the chi-square to degrees of freedom was approximately 3, indicating a moderate fit

    (Bentler & Bonett, 1980). The three factor model fit significantly better than the next-best fitting

    model (2a), with a chi-square difference of 54.82 (df = 2), significant at the p .001 level. The

  • Multidimensional Mentoring Measure 15

    fit indices indicated an excellent fit, with the SRMSR = .046, GFI = .92, TLI = .93, NFI = .94,

    and CFI = .95. (We recognize that since this is the sample used to modify the measure, a good fit

    was expected.) However, as with the 15-item measure, the RMSEA was .11 (above the .10

    recommendation; Browne and Cudeck, 1993), and the RMSEA confidence interval overlapped

    with model 2a.

    ----------------------------------

    Insert Table 5 about here

    ----------------------------------

    The completely standardized factor loadings and error variances for the MFQ-9 are

    presented in Table 6. All of the loadings were statistically significant (p .01), and most were

    large, ranging from .69 to .89 with an average factor loading of .79.

    Disattenuated correlations and average variance indices are presented in Table 7.

    Correlations ranged from .76 to .80, with an average of .78. The correlations were significant (p

    .01) and significantly less than 1.0 (p .01). Although high, the correlations between the

    subdimensions were not high enough to suggest a second order factor. Finally, the hypothesized

    dimensions accounted for a large proportion of variance (on average) in their indicators.

    Specifically, vc() = .58 for career support, vc() = .65 for psychosocial support, and vc() = .63

    for role modeling. Comparing vc() with the disattenuated correlations, four of the six

    comparisons fit the criteria, and the other two were close. The vc() for the career support scale

    was .58, which was slightly less than the square of the disattenuated correlation between career

    support and role modeling (.61). Additionally, vc() for the role modeling scale was .63, which

    was less than the square of the disattenuated correlation between psychosocial support and role

    modeling (.64).

  • Multidimensional Mentoring Measure 16

    ----------------------------------------------

    Insert Tables 6 and 7 about here

    ----------------------------------------------

    The correlation between the MFQ-9 and Ragins and Mc Farlins (1990) 27-item measure

    (excluding the social and parent scales) was r = .73. This correlation is identical to the

    correlation between the 15-item measure and Ragins and Mc Farlins (1990) 27-item measure.

    Reliability and item-total correlations. The coefficient alpha internal consistency

    reliability for the scores on the MFQ-9 was .91. The reliabilities for the scores on the career

    support scale, psychosocial support scale, and role modeling scale were .82, .85, and .83. Item-

    total correlations for all three scales items ranged from .62 to .78. Item analyses indicated that

    the deletion of any single item would not result in improvement in these reliabilities.

    Comparison of Ragins and McFarlins Measure and the Revised 9-Item Measure

    Using hierarchical regression, the ability of the two measures to explain variance in

    outcomes over the other was evaluated (cf. Liden & Maslyn, 1998). The results are presented in

    Table 8. The addition of either measure to equations with the other did not explain statistically

    significant increases in variance in the outcome measures.

    ----------------------------------------

    Insert Table 8 about here

    ----------------------------------------

    Assessment of Concurrent Validity

    Table 9 presents the means, standard deviations, and intercorrelations for the measures.

    The 15-item Scandura and Ragins measure and the MFQ-9 were highly correlated (r = .94), and

    outcome variable correlations with the MFQ-9 were very similar to those with the 15-item

    measure. The MFQ-9 was significantly related to job satisfaction (r = .23, p .01) and

    marginally significantly related to organizational commitment (r = .14, p .10). The career

  • Multidimensional Mentoring Measure 17

    support and role modeling dimensions were significantly related to satisfaction (r = .20 and .26, p

    .05 and p .01, respectively) and commitment (r = .13 and .19, p .10 and p .05,

    respectively). Psychosocial support was significantly related to satisfaction (r = .15, p .10) but

    not to commitment (r = .06). None of the three scales were significantly related to anxiety.

    (Parenthetically, the statistical significance of the correlations between the MFQ-9 and the three

    outcome variables are identical to the statistical significance of the same correlations for the

    Ragins and McFarlin [1990] measure.)

    -------------------------------------

    Insert Table 9 about here

    -------------------------------------

    Discussion

    The initial findings of Study 2 offered only moderate support for the validity of the 15-

    item measure. Using the construct definitions as well as the empirical results of Studies 1 and 2,

    we revised the measure to 9 items and reevaluated it, repeating the analyses. Substantial support

    was obtained for the MFQ-9. However, this is not surprising, since the same data used to revise

    the measure were used to reevaluate it. A stronger test comes from an additional, confirmatory

    sample. Hence, we conducted Study 3.

    Study 3

    This study was conducted to evaluate the validity of the MFQ-9. Convergent and

    discriminant validity, reliabilities, and item total correlations were evaluated as they were in

    Study 2.

  • Multidimensional Mentoring Measure 18

    Method

    Sample

    This sample was drawn from employed CPAs, and a total of 1024 surveys were

    completed. Of these, 795 indicated having had a mentor at some point in their careers, and thus

    these comprised the study sample. Sixty-eight percent of the mentored sample was male, and the

    average age was 30. Twelve percent of the sample was African American, and 88% were

    Hispanic or Caucasian. Average organizational tenure was 2.3 years, ranging from a few months

    to 28 years.

    Measures

    The MFQ-9, drawn from Scandura and Ragins (1993), was used. The reliability for this

    data on the MFQ-9 was .78, and the reliabilities were .77, .67, and .69 for the psychosocial

    support, career support, and role modeling scale data, respectively.

    Analyses

    Convergent and discriminant validity. The convergent and discriminant validity of the

    measure was evaluated using maximum likelihood estimation in Lisrel 8.51 (Jreskog &

    Srbom, 1993). Sample covariances served as input for the estimations, each item was specified

    as loading on only one factor, the errors were specified as uncorrelated among themselves, and

    the latent trait factor correlations were freely estimated. The same four rival models tested in

    Study 2 were tested here, and the same examination of convergent and discriminant validity was

    conducted.

    Scale reliabilities and item-total correlations. Coefficient alpha internal consistency

    reliabilities and item-total correlations were examined for the MFQ-9 and its scales.

  • Multidimensional Mentoring Measure 19

    Results

    Convergent and Discriminant Validity

    The fit statistics for the four alternative models are presented in Table 5. The

    hypothesized three factor model fit best, with 2 = 61.07 and df = 24 (p .01), and a ratio of the

    chi-square to degrees of freedom of approximately 2, indicating a good fit (Bentler & Bonett,

    1980). Additionally, the three factor model fit significantly better than the next-best fitting

    model (2a), with a chi-square difference of 215.41 (df = 2), significant at the p .001 level.

    Finally, the fit indices indicated an excellent fit: SRMSR = .031, GFI = .98, TLI = .96, NFI = .96,

    CFI = .98, and RMSEA = .04. The RMSEA confidence interval comparisons show that the three

    factor model is significantly better than any of the others (i.e., there is no overlap with any of the

    other models confidence intervals).

    The completely standardized factor loadings and error variances for the MFQ-9 are

    presented in Table 6. All of the loadings were statistically significant (p .01), and most were

    large. The loadings ranged from .45 to .81. Four loadings were below .70: item 1 (.68), item 5

    (.45), item 12 (.58), and item 15 (.67).

    Disattenuated correlations and average variance indices are presented in Table 7.

    Correlations ranged from .42 to .54, with an average of .49. The correlations were significant (p

    .01) and significantly less than 1.0 (p .01). The correlations between the subdimensions were

    not high enough to suggest a second order factor. Finally, the hypothesized dimensions

    accounted for a large proportion of variance (on average) in their indicators: vc() = .44 for

    career support, vc() = .55 for psychosocial support, and vc() = .44 for role modeling.

    Comparing vc() with the disattenuated correlations, all twelve comparisons fit Fornell and

  • Multidimensional Mentoring Measure 20

    Larckers (1981) criteria (i.e., the average variance was greater than the square of the constructs

    correlations with other variables).

    Reliability and Item-total Correlations

    The reliability estimate for the data on the MFQ-9 was .78 (reliabilities were .67, .77, and

    .69 for career, psychosocial, and role modeling scale data, respectively). Item-total correlations

    for all three scales items ranged from .38 to .66. Only the deletion of item 5 (rit = .38) in the

    career support scale would result in an improvement in the coefficient alpha (from .67 to .70).

    Discussion

    While the construct validity evidence collected in Study 3 was generally supportive, some

    was not. The fit statistics obtained for the theoretical model indicated an excellent fit of the

    model to the data. Notably, both model comparison techniques (change in 2 and RMSEA

    confidence interval comparisons) supported the three factor model as significantly better than the

    alternative models. Additionally, all of the factor loadings were statistically significant.

    Evidence for discriminant validity was also strong. Each of the factor correlations was

    significantly less than 1.0, and the average variance comparisons using vc() met the criteria in all

    of the comparisons.

    On the other hand, some of the evidence was not so supportive. Four of the factor

    loadings in this study were below the .70 threshold, albeit most only slightly so. Reliability for

    the overall measure was adequate, but two of the scales were slightly below the .70 rule-of-

    thumb. While these results are somewhat surprising, they should not generate too much alarm.

    First, it is important to recall that all of the items demonstrated strong content adequacy (Study

    1). Additionally, the low loadings were not replicated in Study 2--only item 1 had low loadings

  • Multidimensional Mentoring Measure 21

    in both Study 2 and Study 3. Finally, it should be noted that the loadings for item 1 were close to

    .70 (.69 and .68 in Studies 2 and 3).

    The results of this study draw attention to item 5. This item was the only one for which

    deletion would have resulted in an improvement in a scales reliability. Item 5 also obtained the

    lowest factor loading in this study. Given these results, item 5 should probably be looked at

    more carefully in future studies. However, we note that the poor results obtained in this study are

    possibly sample specific, as item 5 was not problematic in Study 2.

    General Discussion

    Unfortunately, as noted by many others (e.g., Cook et al., 1981; Podsakoff & Dalton,

    1987; Schriesheim et al., 1993; Stone-Romero, 1994), the field generally does not devote enough

    attention to measurement development and validation. Researchers tend to use measures to

    evaluate substantive hypotheses prior to evaluating the construct validity of the measures. Thus,

    the purpose of this study was to evaluate the validity of a commonly used measure in mentoring

    research, Scandura and Ragins (1993) mentoring measure.

    The analyses of Scandura and Ragins (1993) 15-item scale in Study 2 were not as

    supportive as expected. While the content validity analysis was excellent (three factors, strong

    theoretical factor loadings, no substantial cross-loadings), the results of some of the other

    analyses were disappointing. The confirmatory factor model did not fit well, in that three of the

    four fit indices employed were below the .90 criterion. Additionally, the comparisons

    recommended by Fornell and Larcker (1981) were not supportive of the distinctiveness of the

    constructs. The poor results of the confirmatory factor analysis and the Fornell and Larcker

    (1981) average variance comparisons suggested that the Scandura and Ragins measure could be

    modified to better represent the construct. The content adequacy results implied that the current

  • Multidimensional Mentoring Measure 22

    items did indeed represent the theoretical constructs, and thus that new items were probably not

    needed. Using the results of the above analyses and the theoretical construct definitions, the

    MFQ-9 was constructed and reevaluated.

    The analyses of the MFQ-9 were very supportive of its validity. Convergent and

    discriminant analyses were strongly supportive of the MFQ-9. In the confirmatory factor

    analyses, the three factor model fit the data extremely well in both Study 2 and Study 3. Factor

    loadings were strong in both samples, with some minor exceptions in Study 3. Similarly,

    reliabilities and item-total correlations were strong in Study 2. While two of the estimated

    reliabilities in Study 3 were below the .70 recommendation (Nunnally & Bernstein, 1994), they

    were only slightly below. These lower factor loadings and reliability estimates could be due to

    the sample (accountants), or to the diverse activities represented by these items since it is a short

    form attempting to capture broad behavioral categories.

    Additional evidence of the convergent and discriminant validity strongly supported the

    MFQ-9. The average variance explained by each latent variable was considerable. Furthermore,

    the comparisons of the squared correlations with the average variance indicator (Fornell &

    Larcker, 1981) all exceeded the criteria in Study 3, supporting the distinctiveness of the factors.

    The correlation between the 15-item scale and the MFQ-9 scale was very high (r = .94),

    indicative that these two measures assess the same global construct. The hierarchical regression

    results also supported the convergent validity of the MFQ-9. It appears to assess the same global

    mentoring construct as Ragins and McFarlins measure. Based on these results, we consider the

    convergent and discriminant validity of the MFQ-9 to be strongly supported. Furthermore, the

    concurrent validity assessment also supported the MFQ-9. All of the relationships were as

    predicted, with one nonsignificant exception (psychosocial support and commitment).

  • Multidimensional Mentoring Measure 23

    A recent study by Pellegrini and Scandura (2005) provided additional support for the

    MFQ-9. A three factor structure was supported in two separate samples (those satisfied with

    their mentors and those dissatisfied with their mentors). Interestingly, the psychometric

    properties of the MFQ-9 were stronger in the dissatisfied sample than in the satisfied sample.

    The authors recommended that further investigation is needed.

    Limitations

    The primary limitation of this study was the use of self-report measures, which may lead

    to method bias. However, as Allen et al.s (2004) recent meta-analysis noted, affective variables

    have stronger correlations with mentoring functions than objective variables (e.g., compensation,

    promotions, salary growth). With the use of affective variables (as in this study), it is necessary

    to assess employees perceptions directly. Additionally, the inclusion of the anxiety measure and

    the differential correlations it obtained with the mentoring scales (as opposed to the positive

    correlations with job satisfaction and organizational commitment) help to alleviate some of the

    concerns of common method bias. That is, it does not appear that respondents were merely

    responding positively to each item.

    A second limitation of this study is the sample in Study 3 used to confirm the factor

    structure of the revised Scandura and Ragins (1993) measure. This sample, comprised entirely of

    CPAs, could be considered unique and the results from analyses potentially ungeneralizable to

    other organizational populations. However, the Study 3 sample was merely used to confirm

    results obtained with a much more diverse sample (Study 2). Relatedly, it would have been ideal

    to evaluate concurrent validity by looking at the relationship of the MFQ-9 with outcome

    variables in Study 3. Unfortunately, we were unable to collect this additional data due to

    organizational constraints.

  • Multidimensional Mentoring Measure 24

    Future Research

    Obviously, given the results of Studies 2 and 3, more research is needed on the validity of

    the MFQ-9. In particular, investigations of relationships with other variables in other

    organizational settings would be useful. Also, further investigation of item 5 is required.

    Additionally, the measurement model tests conducted in Study 2 point to a potentially

    interesting area for further study. Traditionally, role modeling is considered part of the

    psychosocial function (Kram, 1985). Scanduras measure (Scandura, 1992; Scandura and

    Ragins, 1993) proposes that the role modeling function is separate and distinct from the

    psychosocial function. While the analyses in Study 3 supported this distinction, the analyses in

    Study 2 did not. What is particularly interesting is that the Study 2 analyses indicated that role

    modeling was more closely associated with the career support function (not the psychosocial

    function). This presents a potentially useful area for future study, both from a theoretical and an

    empirical standpoint.

    Summary

    The analyses reported in this paper indicated that the original 15-item multidimensional

    mentoring measure (Scandura & Ragins, 1993) needed refinement. Based on the results of

    multiple samples and analyses, the measure was refined to a 9-item measure assessing three

    dimensions: career support, psychological support, and role modeling. This shorter measure still

    provides researchers the ability to look at either the three dimensions separately, or at the

    global mentoring construct. Ultimately, a shorter measure with superior evidence of validity

    had been developed: the 9-item Mentoring Functions Questionnaire (MFQ-9). The MFQ-9 is

    now recommended (instead of the original 15-item version developed by Scandura and Ragins

    [1993]).

  • Multidimensional Mentoring Measure 25

    Footnote

    1 The analyses were also conducted using the full 33-item measure, and the results are

    not substantially different from the analyses with the 27-item measure. The results are available

    from the first author upon request.

    2 The third logical two factor model (with career support and psychosocial support

    loading on one factor and role modeling on the second) was not tested due to theoretical

    considerations. Kram (1985) hypothesized two dimensions, career support and psychosocial

    support, with role modeling a subdimension of psychosocial support. Yet the correlation in the

    Study 2 data indicated that role modeling was more strongly related to career support. Thus,

    Models 2a and 2b were tested to allow evaluation of whether role modeling is better viewed as a

    subdimension of career support or psychosocial support. Comparison with the three factor model

    then allowed determination of whether two or three factors best represented the data.

  • Multidimensional Mentoring Measure 26

    References

    Allen, T.D., Eby, L.T., Poteet, M.L., Lentz, E., & Lima, L. (2004). Career benefits

    associated with mentoring for protgs: A meta-analysis. Journal of Applied Psychology, 89,

    127-136.

    Bentler, P.M., & Bonett, D.G. (1980). Significance tests and goodness-of-fit in the

    analysis of covariance structures. Psychological Bulletin, 88, 588-606.

    Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A.

    Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136162). Thousand Oaks,

    CA: Sage.

    Cook, J.D., Hepworth, S.J., Wall, T.D., & Warr, P.B. (1981). The experience of work.

    London: Academic Press.

    Ensher, E.A., & Murphy, S.E. (1997). Effects of race, gender, perceived similarity, and

    contact on mentor relationships. Journal of Vocational Behavior, 50, 460-481.

    Ford, J.K., MacCallum, R.C., & Tait, M. (1986). The application of exploratory factor

    analysis in applied psychology: A critical review and analysis. Personnel Psychology, 39, 291-

    314.

    Fornell, C., & Larcker, D.F. (1981). Evaluating structural equation models with

    unobservable variables and measurement error. Journal of Marketing Research, 18, 39-50.

    Jreskog, K.G., & Srbom, D. (1993). LISREL 8 user's reference guide. Chicago:

    Scientific Software International.

    Korman, A.K. (1974). Contingency approaches to leadership: An overview. In J. G. Hunt

    & L. L. Larson (Eds.), Contingency approaches to leadership (pp. 189198). Carbondale:

    Southern Illinois University Press

  • Multidimensional Mentoring Measure 27

    Kram, K.E. (1985). Mentoring at work: Developmental relationships in organizational

    life. Glenview, IL: Scott Foresman.

    Lankau, M.J., & Scandura, T.A. (2002). An investigation of personal learning in

    mentoring relationships: Content, antecedents and consequences. Academy of Management

    Journal, 45, 779-790.

    Liden, R.C., & Maslyn, J.M. (1998). Multidimensionality of leader-member exchange:

    An empirical assessment through scale development. Journal of Management, 24, 43-72.

    Mathieu, J., & Zajac, D.M. (1990). A review and meta-analysis of the antecedents,

    correlates, and consequences of organizational commitment. Psychological Bulletin, 108, 171-

    194.

    Medsker, G.J., Williams, L., & Holahan. (1994). A review of current practices for

    evaluating causal models in organizational behavior and human resources management research.

    Journal of Management, 20, 439-464.

    Neilson, T.R., Carlson, D.S., & Lankau, M.J. (2001). The supportive mentor as a means

    of reducing work-family conflict. Journal of Vocational Behavior, 59, 364-381.

    Noe, R.A. (1988). An investigation of the determinants of successful assigned

    mentoring relationships. Personnel Psychology, 41, 457-479.

    Nunnally, J.C., & Bernstein, I.H. (1994). Psychometric theory (3rd

    edition). New York:

    McGraw-Hill.

    Pellegrini, E.K., & Scandura, T.A. (2005). Construct equivalence across groups: An

    unexplored issue in mentoring research. Educational and Psychological Measurement, 65, 323-

    335.

  • Multidimensional Mentoring Measure 28

    Podsakoff, P.M., & Dalton, D.R. (1987). Research methodology in organizational studies.

    Journal of Management, 13, 419-441.

    Porter, L.W., Steers, R.M., Mowday, R.T., & Boulian, P.V. (1974). Organizational

    commitment, job satisfaction, and turnover among psychiatric technicians. Journal of Applied

    Psychology, 59, 603-609.

    Ragins, B.R., & Cotton, J.L. (1999). Mentor functions and outcomes: A comparison o

    men and women in formal and informal mentoring relationships. Journal of Applied Psychology,

    84, 529-550.

    Ragins, B.R., & McFarlin, D.B. (1990). Perception of mentor roles in cross-gender

    mentoring relationships. Journal of Vocational Behavior, 37, 321-339.

    Rule, W.R., & Traver, M.D. (1983). Test-retest reliabilities of the State-Trait Anxiety

    Inventory in a stressful social analogue situation. Journal of Personality Assessment, 47, 276-

    277.

    Scandura, T.A. (1992). Mentorship and career mobility: An empirical investigation.

    Journal of Organizational Behavior, 13, 169-174.

    Scandura, T.A., & Ragins, B.R. (1993). The effects of sex and gender role orientation on

    mentorship in male-dominated occupations. Journal of Vocational Behavior, 43, 251-265.

    Schriesheim, C.A., Powers, K.J., Scandura, T.A., Gardiner, C.C., & Lankau, M.J. (1993).

    Improving construct measurement in management research: Comments and a quantitative

    approach for assessing the theoretical content adequacy of paper and pencil survey-type

    instruments. Journal of Management, 19, 385-417.

  • Multidimensional Mentoring Measure 29

    Schwab, D.P. (1980). Construct validity in organizational behavior. In B. Staw & L.

    Cummings (Eds.), Research in organizational behavior (Vol. 2, pp. 3-43). Greenwich CT: JAI

    Press.

    Sherwood, R.D., & Westerback, M.E. (1983). A factor analytic study of the State-Trait

    Anxiety Inventory utilized with preservice elementary teachers. Journal of Research in Science

    and Teaching, 20, 225-229.

    Speilberger, C.D., Gorush, R.L. & Lushene, R.E. (1970). Manual for the state-trait

    anxiety inventory. Palo Alto, CA: Consulting Psychologists Press.

    Stone-Romero, E.F. (1994). Construct validity issues in organizational behavior. In J.

    Greenberg (Ed), Organizational behavior: The state of the science (pp. 155-179). Hillsdale, NJ:

    Erlbaum.

    Tepper, K., Shaffer, B.C., & Tepper, B.J. (1996). Latent structure of mentoring function

    scales. Educational and Psychological Measurement, 56, 848-857.

    Underhill, C.M. (2006). The effectiveness of mentoring programs in corporate settings:

    A meta-analytical review of the literature. Journal of Vocational Behavior, 68, 292-307.

    Wanous, J.P. (1974). A causal-correlational analysis of the job satisfaction and

    performance relationship. Journal of Applied Psychology, 59, 139-144.

    Weiss, D.J., Dawis, R.V., England, G.W., & Lofquist, L.H. (1967). Manual for the

    Minnesota Satisfaction Questionnaire. Industrial Relations Center, University of Minnesota.

    Williams, E.A. (1999). A field experimental investigation of relationship-based

    mentoring training in team settings and team characteristics: Effects on employee attitudes and

    outcomes. Unpublished Doctoral Dissertation, University of Miami.

  • Multidimensional Mentoring Measure 30

    Table 1

    Exploratory Factor Analysis Results for the Content Adequacy Assessment

    ______________________________________________________________________________

    Factor Pattern Coefficients

    Career Psycho- Role

    Item Support social Model

    ______________________________________________________________________________

    1. My mentor takes a personal interest in my career. .55 -.16 .04

    2. My mentor has placed me in important assignments. .68 .10 -.09

    3. My mentor gives me special coaching on the job. .55 .01 .13

    4. My mentor advised me of promotional opportunities. .69 .04 -.06

    5. My mentor helps me coordinate professional goals. .65 -.02 .05

    6. My mentor has devoted special time and consideration to my career. .63 -.14 -.03

    7. I share personal problems with my mentor. .07 -.67 -.03

    8. I socialize with my mentor after work. .01 -.66 -.06

    9. I exchange confidences with my mentor. .09 -.67 -.06

    10. I consider my mentor to be a friend. -.03 -.67 .10

    11. I often go to lunch with my mentor. -.01 -.56 .02

    12. I try to model my behavior after my mentor. -.13 -.01 .65

    13. I admire my mentors ability to motivate others. -.05 -.17 .47

    14. I respect my mentors knowledge of the profession. .23 .15 .57

    15. I respect my mentors ability to teach others. .09 .11 .61

    ______________________________________________________________________________

    Note. Factor pattern coefficients .40 are underlined.

  • Multidimensional Mentoring Measure 31

    Table 2

    Fit Statistics for Alternative Theoretical Models for Scandura and Ragins Mentoring Measure

    ___________________________________________________________________________

    Confidence

    Factors 2 df SRMSR GFI TLI NFI CFI RMSEA Interval

    ___________________________________________________________________________

    1 510.35 90 0.083 0.73 0.77 0.77 0.80 .16 (.15 ; .17)

    2a 363.98 89 0.072 0.81 0.84 0.83 0.87 .12 (.11 ; .14)

    2b 422.77 89 0.077 0.77 0.81 0.81 0.84 .15 (.14 ; .16)

    3 297.82 87 0.067 0.84 0.88 0.86 0.90 .11 (.10 ; .12)

    ___________________________________________________________________________

    Note. Model 2a allowed the role modeling items to load on the career support factor; model 2b allowed the role modeling items to

    load on the psychosocial support factor. SRMSR = standardized root mean square residual; GFI = goodness of fit index; TLI =

    Tucker-Lewis Index; NFI = normed fit index; CFI = comparative fit index; RMSEA = root mean square error of approximation.

  • Mentoring Measures 32 -

    Table 3

    Completely Standardized Factor Loadings for the Confirmatory Factor Analysis of the

    15-Item Measure

    ______________________________________________________________________________

    Factor Loadings Error Variance

    Scales and Items 1 2 3 (Theta Delta)

    ______________________________________________________________________________

    Career Support

    1. My mentor takes a personal interest in my career. .72 .49

    2. My mentor has placed me in important assignments. .71 .50

    3. My mentor gives me special coaching on the job. .69 .52

    4. My mentor advised me of promotional opportunities. .73 .47

    5. My mentor helps me coordinate professional goals. .71 .49

    6. My mentor has devoted special time and

    consideration to my career. .77 .40

    Psychosocial Support

    7. I share personal problems with my mentor. .77 .40

    8. I socialize with my mentor after work. .72 .48

    9. I exchange confidences with my mentor. .87 .25

    10. I consider my mentor to be a friend. .80 .36

    11. I often go to lunch with my mentor. .63 .60

  • Mentoring Measures 33 -

    Table 3 Continued

    Factor Loadings Error Variance

    Scales and Items 1 2 3 (Theta Delta)

    ______________________________________________________________________________

    Role Modeling

    12. I try to model my behavior after my mentor. .71 .49

    13. I admire my mentors ability to motivate others. .88 .23

    14. I respect my mentors knowledge of the profession. .58 .67

    15. I respect my mentors ability to teach others. .79 .38

    ______________________________________________________________________________

    Note. All parameter estimates are statistically significant (p < .01).

  • Mentoring Measures 34 -

    Table 4

    Average Variance and Disattentuated Correlations for Scandura and Ragins 15-Item

    Measure

    ___________________________________________________________________________

    vc() Factor Intercorrelations (Phi Matrix)

    ___________________________________________________________________________

    1 2 3

    Career Support .52 1.00

    Psychosocial Support .58 .74 1.00

    Role Modeling .56 .83 .74 1.00

    ___________________________________________________________________________

    Note. Correlations are statistically significant (p < .001).

  • Mentoring Measures 35 -

    Table 5

    Fit Statistics for Alternative Theoretical Models for the MFQ-9 Measure

    _____________________________________________________________________________

    Confidence

    Factors Study 2 df SRMSR GFI TLI NFI CFI RMSEA Interval

    _____________________________________________________________________________

    1 2 243.08 27 0.086 .80 .72 .77 .79 .19 (.17 ; .21)

    3 501.54 27 0.090 .86 .58 .67 .68 .16 (.15 ; .17)

    2a 2 134.12 26 0.068 .88 .86 .88 .90 .14 (.12 ; .16)

    3 276.48 26 0.068 .92 .77 .82 .82 .12 (.11 ; .13)

    2b 2 211.20 26 0.082 .82 .75 .80 .82 .19 (.17 ; .21)

    3 298.05 26 0.077 .91 .75 .80 .82 .12 (.11 ; .13)

    3 2 79.30 24 0.046 .92 .93 .94 .95 .11 (.08 ; .13)

    3 61.07 24 0.031 .98 .96 .96 .98 .04 (.03 ; .06)

    ____________________________________________________________________________

    Note. Model 2a allowed the role modeling items to load on the career support factor; model 2b

    allowed the role modeling items to load on the psychosocial support factor. SRMSR =

    standardized root mean square residual; GFI = goodness of fit index; TLI = Tucker-Lewis Index;

    NFI = normed fit index; CFI = comparative fit index; RMSEA = root mean square error of

    approximation.

  • Mentoring Measures 36 -

    Table 6

    Completely Standardized Factor Loadings for the Confirmatory Factor Analysis of the MFQ-9

    Measure

    ______________________________________________________________________________

    Factor Loadings Error Variance

    Scales and Items 1 2 3 (Theta Delta)

    ______________________________________________________________________________

    Career Support

    1. My mentor takes a personal interest in my career. .69(.68) .52(.53)

    5. My mentor helps me coordinate professional goals. .75(.45) .44(.80)

    6. My mentor has devoted special time and

    consideration to my career. .84(.81) .30(.34)

    Psychosocial Support

    7. I share personal problems with my mentor. .74(.71) .46(.49)

    9. I exchange confidences with my mentor. .86(.79) .27(.38)

    10. I consider my mentor to be a friend. .83(.73) .32(.47)

    Role Modeling

    12. I try to model my behavior after my mentor. .73(.58) .47(.66)

    13. I admire my mentors ability to motivate others. .89(.73) .21(.45)

    15. I respect my mentors ability to teach others. .76(.67) .43(.55)

    ______________________________________________________________________________

    Note. Estimates from Study 3 are in parentheses. All parameter estimates are statistically

    significant (p < .01).

  • Mentoring Measures 37 -

    Table 7

    Average Variance and Disattentuated Correlations for the MFQ-9 Measure

    ______________________________________________________________________________

    vc() Factor Intercorrelations (Phi Matrix)

    ______________________________________________________________________________

    1 2 3

    Career Support .58(.44) 1.00

    Psychosocial Support .65(.55) .76(.54) 1.00

    Role Modeling .63(.44) .78(.42) .80(.50) 1.00

    ______________________________________________________________________________

    Note. Estimates from Study 3 are in parentheses. Correlations are statistically significant (p <

    .001).

  • Mentoring Measures 38 -

    Table 8

    Hierarchical Regression Results for the MFQ-9 Measure and Ragins and

    McFarlins 27-Item Measure

    ______________________________________________________________________________

    Dependent Variable,

    Step, and

    Variable Standardized Regression Coefficients

    Entered Step 1 Step 2 R2 R

    2

    Job Satisfaction

    1. MFQ-9 .22** .19 .05** --

    2. Ragins and McFarlin -- .04 .05* .00

    ----------------------------------------------------------------------------------------------------------------------

    1. Ragins and McFarlin .18* .04 .03* --

    2. MFQ-9 -- .19 .05* .02

    ______________________________________________________________________________

    Commitment

    1. MFQ-9 .14 .10 .02 --

    2. Ragins and McFarlin -- .06 .02 .00

    ----------------------------------------------------------------------------------------------------------------------

    1. Ragins and McFarlin .13 .06 .02 --

    2. MFQ-9 -- .10 .02 .00

    ______________________________________________________________________________

    Anxiety

    1. MFQ-9 .03 .18 .00 --

    2. Ragins and McFarlin -- -.20 .02 .02

    ----------------------------------------------------------------------------------------------------------------------

    1. Ragins and McFarlin -.07 -.20 .01 --

    2. MFQ-9 -- .18 .02 .01

    ______________________________________________________________________________

  • Mentoring Measures 39 -

    Table 9

    Correlations of Mentoring Measures and Subdimensions with Other Variables (Study 2)

    ____________________________________________________________________________________________________________

    Variable Mean SD 1 2 3 4 5 6 7 8

    ____________________________________________________________________________________________________________

    1. 15-item measure 3.52 .83 (.93)

    2. MFQ-9 3.52 .91 .94** (.91)

    3. Career supporta 3.69 .94 .85** .88** (.82)

    4. Psychosocial supporta 3.26 1.08 .80** .89** .65** (.85)

    5. Role modelinga 3.60 1.04 .86** .90** .68** .68** (.83)

    6. Ragins/McFarlinb 5.16 1.19 .73** .73** .68** .60** .69** (.91)

    7. Job satisfaction 3.62 .74 .28** .23** .20* .15 .26** .18* (.86)

    8. Commitment 3.35 .69 .20* .14 .13 .06 .19* .14 .51** (.85)

    9. Anxiety 2.47 .53 -.01 .03 -.05 .11 .01 -.07 -.37** -.35** (.90)

    ____________________________________________________________________________________________________________

    Note. Coefficient alpha reliabilities are on the diagonal in parentheses. N = 160.

    a Three-item scales from the MFQ-9 measure

    b Ragins and McFarlins 27-item measure of mentoring (excluding the social and parent scales)

    p .10. * p .05. ** p .01.

    University of MiamiScholarly Repository1-1-2004

    Validity of Scandura and Ragins' (1993) Multidimensional Mentoring Measure: An Evaluation and RefinementStephanie L. CastroTerri A. Scandura PhDEthlyn A. WilliamsRecommended Citation