PDFlib PLOP: PDF Linearization, Optimization, Protection
Page inserted by evaluation versionwww.pdflib.com – [email protected]
Journal of Food Processing and Preservation
29
(2005) 208–227.
All Rights Reserved.
208 ©
Copyright 2005, The Author(s)Journal compilation © Copyright 2005, Blackwell Publishing
CONSUMER-BASED OPTIMIZATION OF PEANUT-CHOCOLATE BAR USING RESPONSE SURFACE METHODOLOGY
EDITH M. SAN JUAN
1
, ERMINA V. EDRA
1
, EVANGELINE N. FADRIGALAN
1
, ALICIA O. LUSTRE
1
and ANNA V.A. RESURRECCION
2,3
1
Food Development CenterNational Food Authority
FTI Complex, TaguigMetro Manila, 1632 Philippines
2
Department of Food Science and TechnologyUniversity of Georgia1109 Experiment St.
Griffin, GA 30223-1797
Accepted for Publication July 5, 2005
ABSTRACT
The acceptability of the sensory properties of a peanut-chocolate bar wasoptimized for consumer acceptance using response surface methodology. Thefactors studied included sugar, peanuts, cocoa powder and a process variable,degree of roast. Twenty-seven peanut-chocolate bar formulations with tworeplications were evaluated for consumer acceptance (
n
=
168) for overallliking and acceptance of color, appearance, flavor, sweetness and textureusing 9-point hedonic scales. In terms of overall liking, the use of dark-roastedpeanuts received the largest number of acceptable formulations when com-pared to the medium- and light-roasted peanuts. Sensory evaluation indicatedthat sweetness acceptance was the limiting factor for acceptability. An accept-able peanut-chocolate bar can be obtained by using formulations containing44–54% dark-, medium- or light-roasted peanuts, 1–4% cocoa powder and41–55% sugar.
INTRODUCTION
Peanuts (
Arachis hypogea
) are a highly acceptable food item in thePhilippines (Muego-Gnanasekharan and Resurrecion 1993) and other coun-
Blackwell Publishing Ltd.Oxford, UK and Malden, USAJFPPJournal of Food Processing and Preservation0145-8892Copyright 2005, Blackwell Publishing2005293208227Original Articles
OPTIMIZATION OF PEANUT-CHOCOLATE BAR
E.M. SAN JUAN
ET AL.
3
Corresponding author. TEL: (770) 412-4736; FAX: (770) 412-4748; EMAIL: [email protected]
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 209
tries throughout the world (McWatters 1983). Peanut products manufacturedin the Philippines include peanut butter, candy bars, brittles and confections(Muego-Gnanasekharan and Resurreccion 1993).
A popular peanut confectionery product in the Philippines is a peanut-chocolate bar prepared from a mixture of roasted peanuts, sugar, milk powderand cocoa powder molded into solid form. The peanut-chocolate bar is avail-able in the Philippines under different brands in a variety of shapes includinground and rectangular or bar shapes. The bar-shaped Chocnut is the mostcommonly available brand. Hany and Ricoa Curly Tops are other brands. Mostpeanuts used in the manufacture of peanut-chocolate bars are obtained frommedium-sized peanuts. The utilization of peanuts in a peanut-chocolate barprovides an additional product line and profit to peanut manufacturers. Peanut-chocolate bars in the Philippines exhibit a sweet taste and little peanut aroma.An optimization study on peanut-chocolate bars will be helpful when con-cocting peanut-chocolate bars as a mixture experiment to optimize consumeracceptance of the peanut-chocolate bar.
When formulating peanut-chocolate bar blends, the combinations of thecomponents must always total 100% of the peanut-chocolate bar formulations(Cornell 1982). Response surface designs are used to identify combinationsof a number of experimental variables that will lead to an optimum (Gaculaand Singh 1984). Response surface methodology (RSM) is a system foroptimizing variables by testing several variables at a time, using specialexperimental designs to cut costs, and determining several effects by objectivetests (Henika 1982). RSM uses quantitative data from appropriate experimen-tal designs to determine and simultaneously solve multivariate equations (Gio-vanni 1983). The equations can be used to describe how the test variablesaffect the response, to determine the relationships among the test variablesand to describe the combined effect of test variables on the response.
Mixture RSM is used to systematically evaluate multiple variables whileminimizing the number of evaluations that must be conducted. In mixtureexperiments, the components in the mixture are expressed as a fraction of thetotal mixture, and the response is a function of the proportions of the compo-nents and not the total amount of the mixture (Snee 1974). The componentsin the mixture are subject to two constraints, such that (1) each componentmust contribute a fraction between 0 and 1 to the mixture and (2) the fractionscontributed by each component must equal to 1 or 100% (Hare 1974). Chang-ing the level of one variable will change the level of at least another variablein the experiment. The ratio of the ingredients in the mixture is much moreimportant than the absolute amount of the mixture. In mixture design studies,the components or ingredients are the independent variables.
This study was conducted to optimize a peanut-chocolate bar usingselected proportions of the following components: sugar, peanuts and cocoa
210 E.M. SAN JUAN
ET AL.
powder as well as degree of roast for the peanuts. The specific objectives wereto (1) determine overall liking and acceptability of appearance, color, flavor,sweetness and texture of peanut-chocolate bar which vary as a result of thedegree of roast for peanuts determined by the lightness (
L
value) of peanutsand the levels of sugar, peanuts and cocoa powder in the formulation usingconsumer-affective tests; (2) determine the effects of components onacceptability and (3) identify the levels of sugar, peanuts and cocoa powderas well as the degree of roast that will result in an acceptable peanut-chocolatebar.
MATERIALS AND METHODS
Experimental Design
A three-component constrained simplex lattice mixture design asdescribed by Cornell (1982) was used to optimize the formulation and degreeof roasting of peanuts for the manufacture of peanut-chocolate bars that areacceptable to consumers. Preliminary experiments were conducted to deter-mine the levels at which the components of peanut-chocolate bars could beoptimized. This involves identification of ingredients and levels that areimportant to the acceptance of the product (Schutz 1983). The ingredientsexhibiting the most influence on peanut-chocolate bar quality when alteredwere determined. The most influential ingredients were sugar, peanuts andcocoa powder. Initially, the formulation of a peanut-chocolate bar manufac-turer was used as a basis for varying the levels of the ingredients. Commercialpeanut-chocolate bars with the highest and lowest levels of the ingredientsthat resulted in an acceptable peanut-chocolate bar were prepared. Theseproportions were used as constraints in the mixture experiment where thehighest and lowest levels were identified as the extreme vertices in the con-strained region. On the basis of the components to be studied, nine formula-tions were obtained.
The three mixture components were sugar (
x
1
), peanuts (
x
2
) and cocoapowder (
x
3
) comprising a total of 94.5% of the peanut-chocolate bar formu-lation. The remaining percentage of the peanut-chocolate bar is 5.5% milkpowder, an amount held constant in the formulation. The range of the com-ponents were 45–64% sugar, 35–54% peanuts and 1–4% cocoa powder equiv-alent to a total of 100% of the mixture based on the preliminary experiments.
Three degrees of roast for peanuts were also investigated: light roast (
L
value
=
51), medium roast (
L
value
=
48) and dark roast (
L
value
=
45). Onthe basis of the number of components and degrees of roast studied, 27formulations were obtained.
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 211
In this design, the number of points (
n
) necessary to run a mixtureexperiment is
where
q
is the number of components being studied. Therefore, the minimumnumber of points studied was (2
3
-
1) or 7 (Scheffé 1958) as presented inFig. 1A,B. The seven points are located in four extreme vertices (mixtures 1,2, 3 and 4), two edge centroids (mixtures 5 and 6) and a center point or overallcentroid (mixture 7) (Snee 1974). Two additional peanut-chocolate bar blends(mixtures 8 and 9) were included to provide extra points within the mixturetriangle to support a second-order polynomial. The total number of formula-tions is nine points. A process variable, degree of roast, was selected at three
L
values (
L
values
=
45, 48 and 51). The 27 selected blends or formulationsfor peanut-chocolate bars are presented in Table 1. Two replications of thestudy were conducted.
Preparation of Raw Materials
Raw, shelled medium Florunner type peanuts (
Arachis hypogea L
. 2000crop, Tara Foods, Albany, GA) weighing 12 kg were manually sorted fordamaged kernels and foreign material. The sorted peanuts were divided intotwo 6-kg batches and blanched using a rotary gas roaster (Model L5, Probat,Inc., Memphis, TN) preheated at 204C (400F) and maintained at 101C (214F)for 2.5 min (Plemmons 1997). After heating, the peanuts were air cooled for10 min (Hinds
et al.
1994) in a perforated cooling tray with an inside diameterof 65 cm and a depth of 12 cm. The peanuts were deskinned using a dry peanutblancher (Model EX, Ashton Food Machinery Co., Inc., Newark, NJ). Thepeanuts were manually sorted to remove remaining testa and damaged nuts.Kernels with any remaining testa were passed through the blancher a secondtime. Blanched peanuts were roasted on the basis of the methods of Muego
et al.
(1990) as adapted from Woodroof (1983). Four and a half (4.5) kilo-grams of blanched peanuts were roasted using a rotary gas roaster (with 4.5-kg capacity) preheated at 177C (350F) and maintained at 138C (280F) forapproximately 8, 9 or 10 min for light-, medium- or dark-roasted peanutsequivalent to lightness (
L
values) of 51, 48 and 45, respectively.The exact time of roasting was based on the number of minutes to reach
a Hunter color lightness (
L
) value equivalent to 45 (dark roast), 48 (mediumroast) and 51 (light roast). To monitor the color of peanuts during the roastingprocess, samples were removed every 60 s and measured for color using aGardner Laboratory XL-800 series tristimulus colorimeter with an XL-845circumferential sensor (Pacific Scientific, Bethesda, MD), until the kernelsreached the desired degree of roast. The final
L
value of the peanuts wasmeasured by calculating the mean of four readings.
n q= -2 1
212 E.M. SAN JUAN
ET AL.
FIG. 1. (A) CONSTRAINED REGION IN THE SIMPLEX COORDINATE SYSTEM DEFINED BY THE FOLLOWING RESTRICTIONS: 0.45
£
x
1
£
0.64, 0.35
£
x
2
£
0.54, 0.01
£
x
3
£
0.04 FOR SUGAR, PEANUTS AND COCOA POWDER, RESPECTIVELY. (B) CONSTRAINED REGION
USED IN THE OPTIMIZATION OF PEANUT-CHOCOLATE BARS(A) Formulations used in the mixture experiment are designated as 1–9. (B) Percent composition of
each formulation is listed in parentheses as follows: (sugar, peanuts, cocoa powder).
1 (0.450, 0.540, 0.010)
8 (0.490, 0.485, 0.025)
9 (0.580, 0.395, 0.025)
4 (0.610, 0.350, 0.040)
5 (0.550, 0.440, 0.010)
3 (0.640, 0.350, 0.010)
7 (0.540, 0.435, 0.025)
2 (0.450, 0.510, 0.040)
6 (0.530, 0.430, 0.040)
B
A
Sugar (x1)
Peanut (x2)
Cocoa powder (x3)
1 8
4 5
3
6
2
9 7
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 213
Roasted peanuts were ground once through a Morehouse mill (More-house Industries, Fullerton, CA) set at a stone clearance of 0.25 mm (10notches) and maintained at 77C (171F) using a mixture of steam and water(Muego-Gnanasekharan and Resurreccion 1993). Extra fine granulated sugar(Dixie Crystals, Savannah Foods, Inc., Savannah, GA) was milled into pow-der form by passing through two disc mills. The first pass was through aHammer mill (Horvick Manufacturing, Fargo, ND) with a 3-mm screen size,and the second pass was through a Retsch Mill (Type ZM1, Retsch GmbH,
TABLE 1.COMPOSITION OF PEANUT-CHOCOLATE BAR FORMULATIONS WITH THREE ROASTS USED IN A THREE-COMPONENT CONSTRAINED SIMPLEX LATTICE MIXTURE DESIGN
Formulation number Degree of roast(
L
value)*Component proportion (%)†
Sugar (
x
1
) Peanuts (
x
2
) Cocoa powder (
x
3
)
1 45 45.00 54.00 1.02 45 45.00 51.00 4.03 45 64.00 35.00 1.04 45 61.00 35.00 4.05 45 55.00 44.00 1.06 45 53.00 43.00 4.07 45 54.00 43.50 2.58 45 49.00 48.50 2.59 45 58.00 39.50 2.5
10 48 45.00 54.00 1.011 48 45.00 51.00 4.012 48 64.00 35.00 1.013 48 61.00 35.00 4.014 48 55.00 44.00 1.015 48 53.00 43.00 4.016 48 54.00 43.50 2.517 48 49.00 48.50 2.518 48 58.00 39.50 2.519 51 45.00 54.00 1.020 51 45.00 51.00 4.021 51 64.00 35.00 1.022 51 61.00 35.00 4.023 51 55.00 44.00 1.024 51 53.00 43.00 4.025 51 54.00 43.50 2.526 51 49.00 48.50 2.527 51 58.00 39.50 2.5
*
L
value is the lightness of a color determined with Hunter color lightness,
L
(Anonymous 1979).† The three components total to 94.5% of the peanut-chocolate bar formulation. Milk powder is the
ingredient added in a fixed amount in the different formulations.
214 E.M. SAN JUAN
ET AL.
Haan, West Germany) with 0.25-mm screen size. The milled sugar wasstored at
-
17C in a walk-in freezer until time of use. Two brands of cocoapowder, Brand A made in the Philippines and Brand B imported from Sin-gapore, were thoroughly mixed in a 50:50 w:w ratio. The mixed cocoa pow-der was stored at
-
17C in a walk-in freezer until time of use. Whey milkpowder (Westfarm Foods, Chicago, IL) was stored at
-
17C in a walk-infreezer until time of use.
Processing of Peanut-Chocolate Bars
Twenty-seven peanut-chocolate bar mixtures were prepared by blendingthe ingredients (ground peanuts, sugar and cocoa powder) on the basis of theexperimental design presented in Table 1. Mixture ingredients and milk pow-der were mixed in a Hobart mixer (Model A-200, Troy, OH) for at least 15 minuntil a uniform blend was obtained. The peanut-chocolate bar mixture wasformed into discs, 3-cm diameter
¥
1-cm height (8–10 g) using a hydraulicpress (Carver Laboratory Press, Model M, Menomonee Falls, WI) at a pres-sure of 0 psi and individually wrapped in 10
¥
8 cm precut aluminum foil. Thepeanut-chocolate bars were stored at
-
19C in a walk-in freezer until consumertests were conducted. Before consumer testing, the peanut-chocolate barswere placed in
~
62-g (4 oz) plastic cups with covers and coded using three-digit random numbers.
Sensory Analysis
Consumer tests were conducted in three central locations, Athens, Atlantaand Barnesville, GA. The panelists were recruited on the basis of the followingcriteria: (1) born in the Philippines, (2) aware of no food allergies, (3) betweenthe ages of 18 and 70, (4) satisfied a preselected gender balance requirementof 50% males and 50% females (only one person of each gender per imme-diate family was permitted) and (5) reported eating peanut-chocolate bars orother related products at least 10 times in their entire lifetime.
Design parameters incorporated 25 responses per peanut-chocolate barformulation. In a balanced incomplete block design, each consumer ratedeight of the 54 peanut-chocolate bars (27 formulations
¥
2 replications) andone control peanut-chocolate bar. A total of 168 consumers were required forthe study. Attributes evaluated were overall liking and acceptability of color,appearance, flavor, sweetness and texture using a 9-point hedonic scale where1
=
dislike extremely, 5
=
neither like nor dislike and 9
=
like extremely.“Willingness to buy” was rated using a dichotomous scale with 1
=
“yes,willing to buy” and 2
=
“no, not willing to buy.” A control peanut-chocolatebar, a popular commercial peanut-chocolate bar, was formed into discs to
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 215
obtain a uniform shape similar to the formulated peanut-chocolate bars toprevent stimulus error because of shape.
Tables lined with white paper were placed in an area of an open roomfor sensory panelist evaluation of peanut-chocolate bars. Separate tables wereused by the panelists to complete the demographic questionnaires before thesensory evaluation. The ballots in the preselected order of evaluation werepresented to each panelist. The preselected order of eight peanut-chocolatebar formulations was randomized for each panelist. The panelists wereinstructed to evaluate five peanut-chocolate bar formulations, take a 1-minbreak and evaluate the four remaining peanut-chocolate bar formulations. Thepanelists were asked to place at least 1/4 of the peanut-chocolate bar in theirmouths for evaluation. The panelists were also instructed to drink water afterevery peanut-chocolate bar sample and not to make comments during evalu-ation to prevent influencing the other panelists.
Statistical and Data Analysis
Analysis of variance, using the general linear model procedure (PROCGLM) was conducted on each attribute and “willingness to buy” to determinethe significant differences among peanut-chocolate bars with selected amountsof sugar, peanuts and cocoa powder as well as degree of roast. The model foreach attribute included the main effects of replication: sugar, peanuts, cocoapowder and degree of roast; cross products: sugar
¥
peanuts, sugar
¥
cocoa,peanuts
¥ cocoa, degree of roast ¥ sugar, degree of roast ¥ peanuts and degreeof roast ¥ cocoa; and the square term: degree of roast ¥ degree of roast.Fisher’s least significant difference at P = 0.05 was used to compare the meansof the attributes and “willingness to buy.”
All data were analyzed using Statistical Analysis Software System (SASInstitute, Cary, NC) and procedures in the user’s guide (SAS 1985). Predictionmodels were developed, and models were fitted as described by Cornell(1982). Parameter estimates were determined using regression analysis(PROC REG) with the no intercept (NOINT) option on raw data. The NOINTfunction is used because SAS automatically inserts an intercept. In mixtureexperiments, the limitation of x1 + x2 + x3 = 1.0 requires a NOINT function toobtain the correct parameter estimates. The NOINT function was performedon each dependent variable: overall liking and acceptance of color, appear-ance, flavor, sweetness and texture as well as the following linear independentvariables: degree of roast, sugar, peanuts, cocoa powder; cross product terms:degree of roast ¥ sugar, degree of roast ¥ peanuts, degree of roast ¥ cocoapowder, sugar ¥ peanuts, sugar ¥ cocoa powder, peanuts ¥ cocoa powder; andthe quadratic term: degree of roast ¥ degree of roast. In mixture designs,quadratic terms cannot be included for mixture components; therefore, a
216 E.M. SAN JUAN ET AL.
quadratic term was only included for the processing variable, degree of roast.Response surface models were generated using the second-degree polynomial(Scheffé 1958):
where y = sensory characteristic or response, x1 = degree of roast, x2 = sugar,x3 = peanuts, x4 = cocoa powder and b = parameter estimate for variables usedin the prediction model.
After running PROC REG on all of the dependent variables, SAS indi-cated that the independent variable, degree of roast ¥ cocoa powder, a cross-product term, was biased and inserted a value of zero for the parameterestimates. The cross-product term, degree of roast ¥ cocoa powder, wasremoved from all models and the PROC REG with NOINT option wascalculated again on the remaining independent variables. By removing degreeof roast ¥ cocoa powder from the model, SAS provided parameter estimatesfor all the remaining independent variables for each dependent variable withno bias involved.
The models obtained contain uncorrected values for degrees of freedom(df) and sums of squares (SS). The df is inflated by 1, and the SS for the modelare also inflated. Because the SS for the model and the uncorrected total areinflated, so are the values for the F-test and coefficients of determination (R2).To determine the correct values for the F-test and R2, the constant term (orintercept) was included in the fitted model (Cornell 1990) when running theregression (PROC REG). R2 values were based on the means of responses for25 observations, 27 formulations and each replication. All models withR2 > 0.50 were selected for the development of final models.
The significance of the linear independent variables and cross productswas determined to reduce the models. The stepwise procedure was selectedto retain the significance for independent variable testing at a = 0.15. Linearterms were always retained in the model if a quadratic or cross-product terminvolving that linear term remained in the model. Ideal models are reduced toinclude the least number of terms in the model, making for easier interpreta-tion of the mixture system (Cornell 1981). The F-value was used to determinesignificant differences between the reduced and full models (a = 0.05). Modelsignificance at the 0.05 level was determined using the F-ratio of means squarecalculated as follows (Cornell 1981):
y x x x x x x x x x
x x x x x x x x
= + + + + + ++ + + +b b b b b b b
b b b b1 1 2 2 3 3 4 4 11 1
212 1 2 13 1 3
14 1 4 23 2 3 24 2 4 34 3 4
F =--
¥
Sum of squares full model Sum of square reduced model
Number of terms in full model Number of terms in reduced model
Residual mean square of full model
1
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 217
If no significant difference (a £ 0.05) was identified between the full andreduced models, the reduced model was selected to predict the responsevariable. After the models were reduced, regression (PROC REG) with theNOINT function was performed again on raw data to obtain correct parameterestimates, and PROC REG on the mean data was used to determine correctR2 values. To determine the effects of the mixture components of sugar,peanuts and cocoa powder on the properties of peanut-chocolate bars,response surfaces were generated using PC SAS Graph (SAS 1985).
Attaining the Optimum Formulation
The observation on each design point in sensory evaluation is usuallyrepresented by the mean score of several panelists (Gacula 1993). The datafrom the two replicates were not significantly different from each other andwere combined in the regression analysis. Models with a coefficient of deter-mination (R2) greater than 0.50 (Gills 1998) and significant at P < 0.05 wereused in prediction equations. These were overall liking, color, flavor, sweet-ness and texture, while appearance with R2 < 0.50 was not included in thedevelopment of prediction equations. Contour plots were generated from fullmodels, because the coefficient of determination (R2) of the reduced modelswere low and could not be reduced further.
Prediction models used in the optimization process were obtained fromthe regression analysis using the NOINT option. The acceptable regions onthe contour plot for each dependent variable were defined as formulations thatwere predicted to result in consumer ratings ≥6 (6 = like slightly). The contourplots of the three degrees of roast for each dependent variable were superim-posed to determine the areas of overlap or combinations of the componentsand degree of roast that would result in optimum regions or formulations forpeanut-chocolate bars.
Contour plots were plotted at each degree of roast where the lightness,Hunter L value, corresponding to a specific roast (light, L = 51.0; medium,L = 48.0; dark, L = 45.0), was substituted for the degree of roast variable inthe model. At each degree of roast for each response, there were three contourplots representing each attribute. All three plots were superimposed, and thearea of overlap for all five attributes was considered the optimum region formaximum consumer acceptance.
RESULTS AND DISCUSSION
Modeling Consumer Acceptance of Peanut-Chocolate Bars
The mean consumer ratings for overall liking, “willingness to buy,” color,flavor, appearance, sweetness and texture for peanut-chocolate bars are pre-
218 E.M. SAN JUAN ET AL.
sented in Tables 2 and 3. The mean ratings for overall liking demonstrate thatthe control peanut-chocolate bars exhibited significantly lower mean ratingscompared to the formulated peanut-chocolate bars. Most blends with highsugar contents of 61 or 64% exhibited consumer ratings of <6.0 for overallacceptability. The overall liking, flavor, sweetness and texture attributes of thecontrol peanut-chocolate bars exhibited lower mean consumer ratings, but theconsumer ratings were not significantly different from peanut-chocolate barformulations containing 64% sugar or prepared from peanuts roasted to Lvalues of 48 or 51. The peanut-chocolate bars containing 64% sugar and thehighest degree of roast (formulation 21) were least likely to be purchasedamong the experimental peanut-chocolate bars.
The results of the regression analyses are presented in Table 4 listing thecoefficients of determination (R2) from the “with intercept option” as well asthe prediction models with parameter estimates from the “NOINT option.”Significant models (P < 0.05) having R2 > 0.50 were overall liking, color,flavor, sweetness and texture, while the model for appearance was omitted,because calculations from the model for appearance resulted in an R2 < 0.50.
Figure 2 presents five contour plots of the constrained region for ratingsof overall liking, color, flavor, sweetness and texture obtained using the pre-dictive models for consumer acceptance ratings of the attributes tested. For eachattribute, the predicted responses for each degree of roast (L value = 45, 48 or51) are plotted. The shaded areas represent the area of overlap when attributeresponses to the three degrees of roast are superimposed. These shaded regionson the superimposed plots represent values for consumer acceptance for a par-ticular sensory attribute corresponding to ratings of 6 (like slightly) or greater.
Figure 2 presents contour plots for predicted overall liking at each of thethree degrees of roast. Most of the formulations in the constrained region wereliked by consumers (>6.0). The overall liking of the formulations increasedwith the degree of roast (51 < 48 < 45). The acceptable number of formula-tions was greatest when the degree of roast was high (L value = 45) and leastwhen the degree of roast was low (L value = 51). The sugar content of theformulations also appeared to influence overall liking, wherein formulationswith higher sugar content (61–64%, respectively) exhibited lower ratingsirregardless of the degree of roast.
Most formulations within the constrained area were acceptable in color.The contour plots for color demonstrate that as the degree of roast increased fromlight to dark for peanuts with L values of 51 to peanuts with L values of 45, moreacceptable peanut-chocolate bar formulations were observed. Within the con-strained region, the number of acceptable peanut-chocolate bar formulations forcolor decreased when the color of the peanut-chocolate bars became lighter.
The contour plots for flavor demonstrate that more acceptable formula-tions were observed in peanut-chocolate bars prepared from medium-roasted
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 219
TABLE 2.MEAN CONSUMER RATINGS FOR ACCEPTABILITY OF OVERALL LIKING AND THE
“WILLINGNESS TO BUY” PEANUT-CHOCOLATE BARS*
Formulation Factor levels† Acceptability mean scores
Roast (x1) x2 x3 x4 Overall liking “Willingness to buy”
1 45 0.450 0.540 0.010 6.8 (1.7)ab 1.2 (0.4)g2 45 0.450 0.510 0.040 6.6 (1.8)abc 1.4 (0.9)bcdefg3 45 0.640 0.350 0.010 6.3 (1.7)abcd 1.5 (0.5)abcde4 45 0.610 0.350 0.040 5.9 (2.1)cde 1.6 (0.5)abcd5 45 0.550 0.440 0.010 6.5 (1.7)abc 1.3 (0.5)efg6 45 0.530 0.430 0.040 6.6 (1.7)abc 1.3 (0.5)efg7 45 0.540 0.435 0.025 6.7 (1.4)abc 1.2 (0.4)fg8 45 0.490 0.485 0.025 6.5 (1.7)abc 1.3 (0.5)efg9 45 0.580 0.395 0.025 6.1 (1.9)abcde 1.4 (0.5)cdefg
10 48 0.450 0.540 0.010 6.4 (2.0)abcd 1.4 (0.5)bcdefg11 48 0.450 0.510 0.040 6.7 (1.6)abc 1.3 (0.5)efg12 48 0.640 0.350 0.010 5.6 (2.1)def 1.6 (0.5)abc13 48 0.610 0.350 0.040 6.3 (1.7)abcd 1.4 (0.5)abcdef14 48 0.550 0.440 0.010 7.0 (1.4)a 1.3 (0.5)efg15 48 0.530 0.430 0.040 6.4 (1.6)abcd 1.3 (0.5)efg16 48 0.540 0.435 0.025 6.4 (1.6)abcd 1.3 (0.5)fg17 48 0.490 0.485 0.025 6.3 (1.8)abcd 1.3 (0.4)fg18 48 0.580 0.395 0.025 6.3 (1.6)abcd 1.4 (0.5)cdefg19 51 0.450 0.540 0.010 6.6 (1.7)abc 1.2 (0.4)fg20 51 0.450 0.510 0.040 6.8 (1.6)a 1.2 (0.4)fg21 51 0.640 0.350 0.010 5.4 (2.1)ef 1.7 (0.9)a22 51 0.610 0.350 0.040 5.9 (2.1)bcde 1.4 (0.6)bcdefg23 51 0.550 0.440 0.010 6.1 (1.9)abcde 1.3 (0.5)efg24 51 0.530 0.430 0.040 6.1 (2.0)abcde 1.4 (0.5)cdefg25 51 0.540 0.435 0.025 6.5 (1.7)abc 1.4 (0.5)defg26 51 0.490 0.485 0.025 6.4 (1.7)abcd 1.3 (0.5)efg27 51 0.580 0.395 0.025 6.9 (1.2)a 1.4 (0.5)cdefg28 Control‡ – – – 5.0 (2.0)f 1.6 (0.5)abRange 1.9 0.5
* Each formulation was evaluated by 25 consumers in two replications for a total of 50 responses.Numbers in parentheses refer to a standard deviation of 50 consumer responses per formulation. A9-point hedonic scale was used for acceptability mean ratings (1 = dislike extremely, 5 = neitherlike nor dislike and 9 = like extremely) and a yes or no response for willingness to buy (1 = yesand 2 = no). Mean values in the same column not followed by the same letter are significantlydifferent (P £ 0.05). Range values were calculated as the differences between the highest and lowestmean scores for each dependent variable.
† Factors were the process variable roast (x1) and the proportions of the sugar (x2), peanut (x3) andcocoa powder (x4) components.
‡ Control, commercially available peanut-chocolate bar.
220 E.M. SAN JUAN ET AL.
TAB
LE
3.
ME
AN
CO
NSU
ME
R R
AT
ING
S FO
R A
CC
EPT
AB
ILIT
Y O
F C
OL
OR
, APP
EA
RA
NC
E, F
LA
VO
R, S
WE
ET
NE
SS A
ND
TE
XT
UR
E O
F PE
AN
UT-
CH
OC
OL
AT
E B
AR
S*
Form
ulat
ion
Fact
or l
evel
s†A
ccep
tabi
lity
mea
n sc
ores
Roa
st (
x 1)
x 2x 3
x 4C
olor
App
eara
nce
Flav
orSw
eetn
ess
Text
ure
145
0.45
00.
540
0.01
06.
7 (1
.6)a
6.7
(1.5
)a7.
0 (1
.6)a
6.6
(1.6
)a6.
9 (1
.4)a
245
0.45
00.
510
0.04
06.
4 (1
.7)a
bc6.
5 (1
.6)a
bc6.
5 (1
.8)a
bc6.
4 (1
.8)a
b6.
6 (1
.6)a
bc3
450.
640
0.35
00.
010
6.2
(2.0
)abc
d6.
1 (1
.9)a
bcde
6.1
(1.8
)bcd
e5.
6(2.
1)bc
de6.
0 (2
.0)b
cdef
445
0.61
00.
350
0.04
06.
1 (1
.9)a
bcd
6.2
(1.9
)abc
de5.
7 (2
.0)c
de5.
4 (2
.2)c
de5.
7 (2
.1)c
def
545
0.55
00.
440
0.01
06.
4 (1
.8)a
bcd
6.4
(1.8
)abc
d6.
4 (1
.7)a
bc6.
1 (1
.9)a
bc6.
3 (1
.7)a
bcd
645
0.53
00.
430
0.04
06.
6 (1
.6)a
6.6
(1.6
)abc
6.6
(1.6
)ab
6.3
(1.8
)abc
6.3
(1.7
)abc
d7
450.
540
0.43
50.
025
6.6
(1.4
)a6.
5 (1
.4)a
bc6.
8 (1
.5)a
b6.
6 (1
.6)a
6.7
(1.4
)ab
845
0.49
00.
485
0.02
56.
3 (1
.7)a
bcd
6.4
(1.6
)abc
de6.
3 (1
.8)a
bcd
5.9
(2.2
)abc
d6.
0 (1
.9)b
cdef
945
0.58
00.
395
0.02
56.
0 (1
.8)a
bcd
6.1
(1.6
)abc
de6.
1 (1
.8)b
cde
5.7
(1.8
)abc
de6.
1 (1
.7)a
bcde
1048
0.45
00.
540
0.01
05.
8 (2
.0)b
cd5.
6 (2
.1)e
6.5
(1.8
)abc
6.2
(1.8
)abc
6.1
(1.9
)abc
de11
480.
450
0.51
00.
040
6.6
(1.7
)a6.
5 (1
.7)a
bcd
6.8
(1.5
)ab
6.5
(1.7
)ab
6.6
(1.5
)abc
1248
0.64
00.
350
0.01
05.
7 (2
.1)c
d5.
8 (2
.1)c
de5.
5 (2
.2)d
ef5.
0 (2
.3)e
5.6
(2.1
)def
1348
0.61
00.
350
0.04
06.
4 (1
.5)a
bcd
6.4
(1.4
)abc
d6.
2 (1
.7)a
bcde
6.3
(1.7
)ab
6.2
(1.6
)abc
de14
480.
550
0.44
00.
010
6.7
(1.5
)a6.
7 (1
.5)a
6.7
(1.5
)ab
6.5
( 1.
6)ab
6.9
(1.3
)a15
480.
530
0.43
00.
040
6.5
(1.6
)abc
6.6
(1.5
)abc
6.2
(1.7
)abc
d6.
2 (1
.8)a
bc6.
1 (1
.8)a
bcde
1648
0.54
00.
435
0.02
56.
4 (1
.6)a
bc6.
4 (1
.6)a
bcde
6.3
(1.7
)abc
d6.
3 (1
.8)a
bc6.
5 (1
.4)a
bc17
480.
490
0.48
50.
025
6.4
(1.7
)abc
d6.
4 (1
.6)a
bcd
6.5
(2.0
)abc
6.1
(2.1
)abc
6.4
(1.8
)abc
d
*N
umbe
rs i
n pa
rent
hese
s re
fer
to a
sta
ndar
d de
viat
ion
of 2
5 co
nsum
er r
espo
nses
. Num
bers
in
pare
nthe
ses
refe
r to
a s
tand
ard
devi
atio
n fo
r 50
con
sum
erre
spon
ses
per
form
ulat
ion.
A 9
-poi
nt h
edon
ic s
cale
was
use
d fo
r ac
cept
abili
ty m
ean
ratin
gs (
1 =
disl
ike
extr
emel
y, 5
= n
eith
er li
ke n
or d
islik
e an
d 9
= lik
eex
trem
ely)
. Mea
n va
lues
in
the
sam
e co
lum
n no
t fo
llow
ed b
y th
e sa
me
lette
r ar
e si
gnifi
cant
ly d
iffe
rent
(P
£ 0
.05)
. Ran
ge v
alue
s w
ere
calc
ulat
ed a
s th
edi
ffer
ence
s be
twee
n th
e hi
ghes
t an
d lo
wes
t m
ean
scor
es f
or e
ach
depe
nden
t va
riab
le.
†Fa
ctor
s w
ere
the
proc
ess
vari
able
roa
st (
x 1)
and
the
prop
ortio
ns o
f th
e su
gar
(x2)
, pea
nut
(x3)
and
coc
oa p
owde
r (x
4) c
ompo
nent
s.‡
Con
trol
, com
mer
cial
ly a
vaila
ble
pean
ut-c
hoco
late
bar
.
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 221
1848
0.58
00.
395
0.02
56.
5 (1
.4)a
bc6.
2 (1
.6)a
bcde
6.1
(1.7
)bcd
e6.
1 (1
.6)a
bc6.
2 (1
.6)a
bcde
1951
0.45
00.
540
0.01
06.
1 (1
.9)a
bcd
6.1
(1.9
)abc
de6.
5 (1
.8)a
bc6.
4 (1
.6)a
bc6.
5 (1
.8)a
bc20
510.
450
0.51
00.
040
6.8
(1.6
)a6.
8 (1
.3)a
6.7
(1.5
)ab
6.5
(1.9
)ab
6.7
(1.6
)ab
2151
0.64
00.
350
0.01
05.
6 (2
.1)d
5.7
(2.0
)de
5.4
(2.0
)ef
5.1
(2.0
)de
5.5
(2.1
)ef
2251
0.61
00.
350
0.04
06.
2 (1
.9)a
bcd
6.1
(1.8
)abc
de6.
0 (1
.9)b
cde
5.8
(2.1
)abc
d5.
9 (1
.9)b
cdef
2351
0.55
00.
440
0.01
06.
1 (1
.8)a
bcd
6.4
(1.6
)abc
d6.
2 (1
.8)a
bcde
6.1
(1.9
)abc
6.3
(1.8
)abc
de24
510.
530
0.43
00.
040
6.3
(1.8
)abc
d6.
2 (1
.8)a
bcde
6.1
(2.1
)bcd
e5.
9 (2
.1)a
bcd
6.3
(1.8
)abc
d25
510.
540
0.43
50.
025
6.6
(1.4
)ab
6.4
(1.5
)abc
d6.
4 (1
.7)a
bc6.
1 (1
.7)a
bc6.
6 (1
.5)a
b26
510.
490
0.48
50.
025
6.3
(1.8
)abc
d6.
4 (1
.8)a
bcd
6.5
(1.8
)abc
6.2
(1.9
)abc
6.3
(1.8
)abc
d27
510.
580
0.39
50.
025
6.7
(1.2
)a6.
7 (1
.3)a
b6.
7 (1
.5)a
b6.
3 (1
.6)a
bc6.
6 (1
.6)a
bc28
Con
trol
‡–
––
5.7
(1.8
)cd
5.9
(1.6
)bcd
e4.
9 (2
.2)f
5.0
(2.0
)e5.
2 (1
.9)f
Ran
ge1.
11.
32.
11.
61.
7
Form
ulat
ion
Fact
or l
evel
s†A
ccep
tabi
lity
mea
n sc
ores
Roa
st (
x 1)
x 2x 3
x 4C
olor
App
eara
nce
Flav
orSw
eetn
ess
Text
ure
*N
umbe
rs i
n pa
rent
hese
s re
fer
to a
sta
ndar
d de
viat
ion
of 2
5 co
nsum
er r
espo
nses
. Num
bers
in
pare
nthe
ses
refe
r to
a s
tand
ard
devi
atio
n fo
r 50
con
sum
erre
spon
ses
per
form
ulat
ion.
A 9
-poi
nt h
edon
ic s
cale
was
use
d fo
r ac
cept
abili
ty m
ean
ratin
gs (
1 =
disl
ike
extr
emel
y, 5
= n
eith
er li
ke n
or d
islik
e an
d 9
= lik
eex
trem
ely)
. Mea
n va
lues
in
the
sam
e co
lum
n no
t fo
llow
ed b
y th
e sa
me
lette
r ar
e si
gnifi
cant
ly d
iffe
rent
(P
£ 0
.05)
. Ran
ge v
alue
s w
ere
calc
ulat
ed a
s th
edi
ffer
ence
s be
twee
n th
e hi
ghes
t an
d lo
wes
t m
ean
scor
es f
or e
ach
depe
nden
t va
riab
le.
†Fa
ctor
s w
ere
the
proc
ess
vari
able
roa
st (
x 1)
and
the
prop
ortio
ns o
f th
e su
gar
(x2)
, pea
nut
(x3)
and
coc
oa p
owde
r (x
4) c
ompo
nent
s.‡
Con
trol
, com
mer
cial
ly a
vaila
ble
pean
ut-c
hoco
late
bar
.
TAB
LE
3.
CO
NT
INU
ED
222 E.M. SAN JUAN ET AL.
TAB
LE
4.
PRE
DIC
TIO
N E
QU
AT
ION
S*
FOR
SE
NSO
RY
AT
TR
IBU
TE
S O
VE
RA
LL
AC
CE
PTA
BIL
ITY
AN
D A
CC
EPT
AB
ILIT
Y O
F C
OL
OR
, APP
EA
RA
NC
E,
FLA
VO
R, S
WE
ET
NE
SS A
ND
TE
XT
UR
E
Var
iabl
eM
odel
†R
2
Ove
rall
acce
ptab
ility
2.36
x 1 +
7.2
2x2 -
5.4
8x3 -
410
.86x
4 - 0
.001
1 -
2.4
7x1x
2 - 2
.17x
1x3 +
24.
04x 2
x 3 +
336
.17x
2x4 +
320
.52x
3x4
0.57
33
Col
or a
ccep
tabi
lity
2.89
x 1 -
3.4
5x2 -
5.1
1x3 -
286
.34x
4 - 0
.002
6 -
2.7
2x1x
2 - 2
.72x
1x3 +
33.
06x 2
x 3 +
178
.61x
2x4 +
194
.40x
3x4
0.52
79
App
eara
nce
acce
ptab
ility
0.47
x 1 +
13.
04x 2
+ 1
1.0x
3 + 2
47.7
5x4 +
0.0
060
- 1
.08x
1x2 -
1.1
0x1x
3 + 4
0.82
x 2x 3
- 2
89.2
6x2x
4 - 2
57.9
5x3x
40.
4551
Flav
or a
ccep
tabi
lity
2.60
x 1 +
1.9
3x2 +
8.9
2x3 -
464
.26x
4 - 0
.000
15 -
2.6
4x1x
2 - 2
.71x
1x3 +
20.
25x 2
x 3 +
386
.52x
2x4 +
351
.89x
3x4
0.66
53
Swee
tnes
s ac
cept
abili
ty2.
01x 1
- 1
9.45
x 2 -
13.
34x 3
- 2
11.6
1x4 -
0.0
078
- 1
.29x
1x2 -
1.3
3x1x
3 + 2
7.73
x 2x 3
+ 1
96.9
8x2x
4 + 1
10.4
8x3x
40.
6412
Text
ure
acce
ptab
ility
2.51
x 1 -
3.9
8x2 -
4.7
2x3 -
195
.67x
4 - 0
.002
3 -
2.3
6x1x
2 - 2
.34x
1x3 +
32.
75x 2
x 3 +
103
.72x
2x4 +
96.
72x 3
x 40.
5521
*E
quat
ions
use
d w
ere
the
full
mod
el.
Con
sum
er r
atin
gs b
ased
on
a 9-
poin
t he
doni
c sc
ale
whe
re 1
= d
islik
e ex
trem
ely,
5 =
nei
ther
lik
e no
r di
slik
e an
d9
= lik
e ex
trem
ely.
†x 1
, pro
cess
var
iabl
e ro
ast;
x 2, x
3 an
d x 4
are
the
pro
port
ions
of
the
suga
r, pe
anut
s an
d co
coa
pow
der
com
pone
nts,
res
pect
ivel
y, i
n th
e m
ixtu
re f
or p
eanu
t-ch
ocol
ate
bars
.A
ll m
odel
s si
gnifi
cant
at
P <
0.0
5.
x 12 x 12 x 12 x 12 x 12
x 12
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 223
FIG. 2. CONTOUR PLOTS OF THE CONSTRAINED REGION FOR RATINGS OF OVERALL LIKING, COLOR, FLAVOR, SWEETNESS AND TEXTURE FOR EACH DEGREE OF ROAST
(L VALUE = 45, 48 OR 51) OBTAINED USING PREDICTIVE MODELS FOR CONSUMER ACCEPTANCE RATINGS OF EACH ATTRIBUTE
Shaded areas represent the area of overlap when attribute responses to the three degrees of roast are superimposed. These shaded regions represent consumer acceptance ratings of 6 (like slightly)
or greater.
Roast 45Roast 48
Roast 51Roast 48
Roast 51
Roast 48
Roast 48
Roast 48
Roast 51
Roast 51
Roast 51
Roast 45
Roast 45
Roast 45
5 (55, 44, 1)
1 (45, 54, 1)
Overall liking
Flavor
Texture
Sweetness
Color
2 (45, 51, 4)
8 (49, 48.5, 2.5)
7 (54, 43.5, 2.5)
9 (58, 39.5, 2.5)
4 (61, 35, 4)
3 (64, 35, 1)
5 (55, 44, 1)
3 (64, 35, 1)
4 (61, 35, 4)
9 (58, 39.5, 2.5)
7 (54, 43.5, 2.5)
6 (53, 43, 4)
8 (49, 48.5, 2.5)
1 (45, 54, 1)
3 (64, 35, 1)
3 (64, 35, 1)
3 (64, 35, 1)
4 (61, 35, 4)
4 (61, 35, 4)
4 (61, 35, 4)
6 (53, 43, 4)
6 (53, 43, 4)
6 (53, 43, 4)
2 (45, 51, 4)
2 (45, 51, 4)
2 (45, 51, 4)1 (45, 54, 1)
1 (45, 54, 1)
1 (45, 54, 1)
9 (58, 39.5, 2.5)
9 (58, 39.5, 2.5)
9 (58, 39.5, 2.5)
5 (55, 44, 1)
5 (55, 44, 1)
5 (55, 44, 1)7 (54, 43.5, 2.5)
7 (54, 43.5, 2.5)
7 (54, 43.5, 2.5)
8 (49, 48.5, 2.5)
8 (49, 48.5, 2.5)
8 (49, 48.5, 2.5)
2 (45, 51, 4)
6 (53, 43, 4)
224 E.M. SAN JUAN ET AL.
peanuts with an L value of 48 than in peanut-chocolate bars prepared fromlight-roasted peanuts with an L value of 51. The number of peanut-chocolatebar formulations with acceptable flavor was smallest in chocolate-peanut barsprepared from peanuts roasted to an L value of 51.
A negative effect on the number of acceptable formulations for sweetnesswas observed when more sugar was used in the formulations. With light-roasted peanuts (L = 51), only formulations with less than 55% sugar arepredicted to provide acceptable sweetness. Peanut-chocolate bars preparedfrom peanuts roasted to a medium roast (L = 48) provide the most peanut-chocolate bar formulations acceptable for sweetness; peanuts roasted to a darkroast (L = 45) provide less, whereas peanut-chocolate bars prepared fromlight-roasted peanuts provide the least number of formulations with acceptablesweetness. The sugar content of peanut-chocolate bar formulations providingacceptable sweetness with peanuts roasted to any of the three degrees of roastis approximately 55%.
More peanut-chocolate bar formulations acceptable for texture will beobtained when peanuts are roasted to an L value of 45 or 51 than when peanutsare roasted to an L value of 48. Maintaining sugar concentrations less than55% when preparing peanut-chocolate bars containing peanuts roasted toeither a dark or light roast usually provides an acceptable texture. When usinga medium roast for peanuts, the sugar content may be increased to as muchas 58%, and acceptable textures may be maintained.
Attaining the Optimum Formulation
The optimum formulations and degrees of roast identified during theprocessing of peanut-chocolate bars are presented in the regions of overlap inFig. 3. Figure 3 presents the boundaries of the optimum regions for the threedegrees of roasted peanuts, indicating that acceptance of sweetness is thelimiting factor in the optimization of peanut-chocolate bar formulation. Opti-mum formulations for peanut-chocolate bars contain 44–54% peanuts roastedfrom a light to dark stage, 1–4% cocoa powder and 41–55% sugar totaling100%. When peanuts are medium roasted to an L value of 48, more acceptablepeanut-chocolate bar formulations will be obtained than when roasting to Lvalues of 45 and 51. The number of acceptable formulations increases whensugar content decreases. Formulations with high sugar content (<61%) areacceptable in flavor when medium-roasted (L = 48) peanuts are used. How-ever, sweetness is only acceptable after reduction in sugar content to around56% or lower when combined with medium- to light-roasted peanuts in theformulation of peanut-chocolate bars. A manufacturer wishing to conserveenergy during roasting will presumably choose to formulate peanut-chocolatebars with little sugar and more peanuts. Peanuts grown in geographic locations
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 225
that result in high sugar contents, however, may result in unacceptable peanut-chocolate bar formulations. Peanut maturity and cultivar may also dictatesugar content and result in a deleterious effect on the acceptance of peanut-chocolate bars.
CONCLUSIONS
Mixture RSM was used to determine the effects of variation in concen-trations of sugar, peanuts and cocoa powder as well as degree of roast of
FIG. 3. OPTIMIZED REGIONS OBTAINED BY OVERLAYING CONTOUR PLOTS OF THE CONSTRAINED REGION FOR RATINGS OF OVERALL LIKING, COLOR, FLAVOR,
SWEETNESS AND TEXTURE FOR EACH DEGREE OF ROAST (L VALUE = 45, 48, OR 51)Shaded areas represent areas of overlap for consumer acceptance ratings of 6 (like slightly or greater)
for all attributes.
Roast 45 – color
Roast 48 – color
Roast 51 – color
Roast 45 – texture
Roast 45 – texture
Roast 45 – sweetness
Roast 51 – sweetness
5 (55, 44, 1)
Roast 45 – overall liking
Roast 48 – overall liking
Roast 51 – overall liking
1 (45, 54, 1) 2 (45, 51, 4)
8 (49, 48.5, 2.5)
6 (53, 43, 4)
7 (54, 3.5, 2.5)
9 (58, 59.5, 2.5)
4 (61, 35, 4)
3 (64, 35, 1)
Roast 48 – sweetness
Roast 51 – texture
Roast 45 – flavor
Roast 51 – flavor
Roast 48 – flavor
226 E.M. SAN JUAN ET AL.
peanuts on the sensory attributes of 27 peanut-chocolate bar formulations.Sweetness is the limiting sensory attribute in the manufacture of peanut-chocolate bars. In terms of overall liking, the use of dark-roasted peanutsprovided the greatest number of acceptable peanut-chocolate bar formula-tions, more than either medium- or light-roasted peanuts. Optimum peanut-chocolate bar formulations for any degree of roast were obtained informulations containing 44–54% peanuts, 1–4% cocoa powder and 41–55%sugar.
ACKNOWLEDGMENTS
The authors wish to thank the Peanut-Collaborative Research SupportProgram (P-CRSP) of the United States Agency for International Develop-ment (USAID) Grant No. LAG-G-00-96-90013-00 for providing researchfunds. The opinions herein are those of the authors and do not necessarilyreflect the views of USAID.
REFERENCES
ANONYMOUS. 1979. CIE dimensions for hue and hue-difference and shadesorting with Tektronix 31 and D25-0 Shademaster. Hunterlab ReflectionsTransm. 30, 1–4.
CORNELL, J.A. 1981. Experiments with Mixtures: Designs, Models andAnalysis of Mixture Data, pp. 155–186, John Wiley & Sons, New York,NY.
CORNELL, J.A. 1982. How to Run Mixture Experiments for Product Quality,pp. 1–60, American Society for Quality Control, Milwaukee, WI.
CORNELL, J.A. 1990. Experiments with Mixtures: Design, Models, andAnalysis with Mixture Data, 2nd Ed., pp. 21–98, John Wiley and Sons,Inc., New York, NY.
GACULA, M.C. 1993. Design and Analysis of Sensory Optimization, p. 163,Food and Nutrition Press, Trumbull, CT.
GACULA, M.C. and SINGH, J. 1984. Statistical Methods in Food and Con-sumer Research, p. 214, Academic Press, Inc., Orlando, FL.
GILLS, L.A. 1998. Texture of unstabilized peanut butter and peanut butterstabilized with palm oil. MSc Thesis, University of Georgia, Athens, GA.
GIOVANNI, M. 1983. Response surface methodology and product optimiza-tion. Food Technol. 37, 41–45.
HARE, L.B. 1974. Mixture designs applied to food formulation. Food Tech-nol. 28, 51–56.
OPTIMIZATION OF PEANUT-CHOCOLATE BAR 227
HENIKA, R.G. 1982. Use of response-surface methodology in sensory eval-uation. Food Technol. 11, 96–101.
HINDS, M.J., CHINNAN, M.S. and BEUCHAT, L.R. 1994. Unhydrogenatedpalm oil as a stabilizer for peanut butter. J. Food Sci. 59, 816–820, 832.
MCWATTERS, K.H. 1983. Diversified food uses in peanuts. In Peanuts Pro-duction, Processing, Products, 3rd Ed. (J.G. Woodroof, ed.) p. 309, TheAVI Publishing Co, Inc., Westport, CT.
MUEGO, K.F., RESURRECCION, A.V.A and HUNG, Y.C. 1990. Charac-terization of the textural properties of spreadable peanut based products.J. Texture Studies 21, 61–73.
MUEGO-GNANASEKHARAN, K.F. and RESURRECCION, A.V.A. 1993.Physicochemical and sensory characteristics of peanut paste as affectedby processing conditions. J. Food Process. Pres. 17, 321–336.
PLEMMONS, L.E. 1997. Sensory evaluation methods to improve validity,reliability and interpretation of panelist responses. MS Thesis, Univer-sity of Georgia, Athens, GA.
SAS. 1985. SAS User’s Guide: Basic, Version 5th Ed., SAS Institute, Inc.,Cary, NC.
SCHEFFÉ, H. 1958. Experiments with mixtures. J. Roy. Stat. Soc. B 20, 344–360.
SCHUTZ, H.G. 1983. Multiple regression approach to optimization. FoodTechnol. 11, 46–48, 62.
SNEE, R.D. 1974. Experimental designs for quadratic models in constrainedmixture spaces. Technometrics 17, 149–159.
WOODROOF, J.G. 1983. Peanut butter. In Peanuts Production, Processing,Products, 3rd Ed. (J.G. Woodroof, ed.) pp. 181–225, AVI PublishingCompany, Inc., Westport, CT.