Simulation of GLMM in R
Budget: €8 – €30 EUR
Task 2
- The researcher will perform research where the outcome will be a number variable. In the experiment, he will have two factors (both binary and expected to have interaction in between)
- The experiment will be designed in a way that: - one factor (X1) will be repeated -> every subject (person) will not be exposed to X1 (X1=0) at first and after it will get exposed (X1=1)
- and the second factor X2 will not be repeated and subjects (persons) will be randomly divided into 2 equally sized groups -> X2=0 and X2=1
- The data will be analysed with LMM: y~X1*X2+(1|ID)
- The main aim of the analysis is to show that the effect of factor X2 is bigger at X1=0 than at X1=1. It is expected that the difference is 5. For X2=0 it is expected that effect X1 is 2 (at X2=0, Y on average increases for 2 if X1 increases for 1). For X1=0 it is expected that the effect(influence) X2 equals 2 (at X1=0, Y on average increases for 3 if X2 increases for 1)
- Researcher expects that correlation between 2 consecutive measurements on the same individual is 0.4, -> that the average of variable Y when X1=X2=0 equals 1000, -> that the variance of normally distributed error equals 10, and -> that the coincidence effect is normally distributed (with appropriate variance)
With simulation check, how many participants (approximately) must be included in the research in order to have 80% power and in order to have 90% power? How many participants would you need (both 80 and 90% power) in case we expect the difference is not 5 but 2.5?
I hope that the instructions are clear and understandable and that no information and details were lost in translation.
Thank you
- The researcher will perform research where the outcome will be a number variable. In the experiment, he will have two factors (both binary and expected to have interaction in between)
- The experiment will be designed in a way that: - one factor (X1) will be repeated -> every subject (person) will not be exposed to X1 (X1=0) at first and after it will get exposed (X1=1)
- and the second factor X2 will not be repeated and subjects (persons) will be randomly divided into 2 equally sized groups -> X2=0 and X2=1
- The data will be analysed with LMM: y~X1*X2+(1|ID)
- The main aim of the analysis is to show that the effect of factor X2 is bigger at X1=0 than at X1=1. It is expected that the difference is 5. For X2=0 it is expected that effect X1 is 2 (at X2=0, Y on average increases for 2 if X1 increases for 1). For X1=0 it is expected that the effect(influence) X2 equals 2 (at X1=0, Y on average increases for 3 if X2 increases for 1)
- Researcher expects that correlation between 2 consecutive measurements on the same individual is 0.4, -> that the average of variable Y when X1=X2=0 equals 1000, -> that the variance of normally distributed error equals 10, and -> that the coincidence effect is normally distributed (with appropriate variance)
With simulation check, how many participants (approximately) must be included in the research in order to have 80% power and in order to have 90% power? How many participants would you need (both 80 and 90% power) in case we expect the difference is not 5 but 2.5?
I hope that the instructions are clear and understandable and that no information and details were lost in translation.
Thank you