1 Notes

Click on ‘hide toolbars’ in the bottom-right corner, for better viewing’

Key variables:

  • Key independent variable = experimental condition. Possible conditions…
    • Control: standard instructor training (no VR)
    • Treatment 1: VR training
    • Treatment 2: online training
  • Dependent variable: score on donning and doffing checklist (expected to be approximately on a 1 to 30 scale)

Link to this document: https://rpubs.com/anshulkumar/VRDonDoffDesign

2 Two-group study

2.1 Design flowchart

2.2 Sample size calculation

## 
##      Two-sample t test power calculation 
## 
##               n = 63.76576
##           delta = 2.5
##              sd = 5
##       sig.level = 0.05
##           power = 0.8
##     alternative = two.sided
## 
## NOTE: n is number in *each* group

A power analysis (results above) shows that if we want to be able to detect an effect size (average difference between treatment and control groups) of 2.5 or more (which roughly corresponds to Cohen’s d = 0.5) on the 30-point donning and doffing score, with power of 0.8 and significance level of 0.05, we need 64 participants in each group (~128 participants total who complete all parts of the study).

3 Three-group study

3.1 Design flowchart

3.2 Sample size calculation

3.2.1 Pairwise

We know from the sample size calculation above that if we want to be able to detect an average difference of 2.5 on the donning-doffing 30-point scale between each pair of groups, we need a sample size of 64 in each of the three groups, meaning 192 participants in total who complete the study.

3.2.2 ANOVA

If we to conduct the following hypothesis test:

Null hypothesis: All three group means (of dependent variable) are same

Alternate hypothesis: All three group means (of dependent variable) are not all equal, differing by at least a medium Cohen’s d effect size.

## 
##      Balanced one-way analysis of variance power calculation 
## 
##               k = 3
##               n = 52.3966
##               f = 0.25
##       sig.level = 0.05
##           power = 0.8
## 
## NOTE: n is number in each group

We find that we require a sample size of 53 for each experimental group (159 total participants completing the study).

4 Final sample size recommendation

Written by Nick: An a priori balanced one-way analysis of variance power calculation was performed for sample size estimation based on a Cohen’s medium effect size of f=0.25. With an alpha = 5% and power = 80%, it is estimated that the sample size with this effect size is approximately 53 independent participants in all 3 study groups, N= 159. Therefore, I propose a sample size of N= 200 for the main objective of our study while allowing for minimal attrition rate while controlling for confounding or moderating variables. To determine a significant effect at the 0.01 alpha level, N= 225 (without increasing the population size to control for confounding variables).

Recommendation from Anshul: Begin pilot data collection on all three experimental groups and then re-estimate parameters and likely required sample size once again.