Upload
adele-randall
View
214
Download
0
Embed Size (px)
Citation preview
Caffeine, Sleep and Attention:
Three things we would all like more of.
Blake Colyer&
Max Breidenstein
College students are deprived of sleep and heavy caffeine users (coffee, tea, energy drinks) which affects alertness
Attention and alertness are linked, so…
Background
To determine how sleep and caffeine effect attention levels in a college aged population
Along the way we will examine the possibility of a significant interaction because when tired, caffeine is a go-to for many students
Objective
http://www.speederandearls.com/
3 x 3 Factorial◦ Factor: Hours of Sleep (Fixed)
6 hours, 8 hours, 10 hours◦ Factor: Caffeine Consumption (Fixed)
0 mg, 95 mg, 190 mg 18 individuals randomly assigned to 9
treatment combinations◦ Experiment replicated 4 times
Experiment Design
Yijk = µ + αi + βj + (αβ)ij + eijk i=1,2,3 j=1,2,3 k=1,2
◦ µ = grand mean of attention◦ αi = fixed effect of hours of sleep
Σ αi =0 ,
◦ βj = fixed effect of caffeine consumption Σ βi =0
◦ (αβ)ij = fixed effect of interaction between sleep and caffeine Σ(αβ)ij =0
◦ eijk = experimental error eijk ~ NIID(0, σe2)
Statistical Model
Materials/Facilities◦ 18 homogeneous sleeping facilities
Sound proof rooms with bed◦ 18 Infrared gaze tracking units and computers
Machine that measures infrared reflection from cornea The more time spent viewing the screen, the greater
the attention to the task Wandering eyes mean a wandering mind
Procedure
http://www.tobii.com/en/about-tobii/what-is-eye-tracking/
Methods◦ Subjects arrive night before with staggered
bedtimes for common awakening hour of 8:00AM◦ Sleep, wake up, drink coffee 20 minutes later◦ Read short article whilst machine tracks eye
movement
Procedure (cont.)
Φ = √((rΣΣτij)/(abσ2)) Ho:τij=0 Ha:τij=1.5σ Φ = √((r4.5 σ2 )/(3*3 σ2)) Φ = √ (r / 2)
Power Analysis
r Φ Df2 = ab(r-1)
Power
5 1.58 36 0.76
6 1.73 45 0.823
7 1.87 54 0.925
8 2 63 0.95
Power = 0.90, α=0.05 The ideal replication size per treatment
combination to achieve the given power was 7. Considering we have 9 treatment combinations, that would imply needing 63 subjects in total. When being realistic about cost, it would make more sense to use 18 subjects and replicate the experiment 4 times
Power Analysis