Beyond ANOVA (Part 7.1) {Repeat Measure ANOVA}

[Neeraj Kaushik]

Repeat Measure Anova (RMA)
 
How is it different from ANOVA?
 
So far we've discussed the statistical techniques used for the Descriptive research where we take the cross-sectional data i.e. we collect the data from various respondents/companies for something (e.g. sales) at any given time period.
 
Now we'll discuss the techniques used for the Experimental research where measurement from the same respondent/company is taken for different time periods/treatments. This is also called Before & after Control group design.
 
When we've readings for same parameter in various categories of same variable, we use ANOVA e.g. Sales is given for the car, jeep & truck.
Here we've 2 variables: Sales & Vehicle (3 categories: car, jeep & truck). 
 
But in repeat measure design, we've Sales of truck before & after a particular event (e.g. Deepawali festival). When there're only 2 situations viz. before & after event, we use t-test (If data is normally distributed) or Wilcoxin Matched Pair test (Data is not normal).  
 
But when we've more than 2 situations to compare e.g. Sales is given for the 4 quarters then either we've to use multiple Paired t-tests or  Repeat measure ANOVA (RMA).
 
Using multiple t-tests increasing the chances for Type-I error, hence we are left with the only option of RMA.
 
So net understanding of RMA is that here the same respondent will get the exposure of all treatments, so we've to remember that here the same respondent will have readings for all the columns.
 
More in next post.
 
Happy Learning
Neeraj