Which t-test to use

#1
Hello, I am confused about which t-test should I use in my analysis.

The case:
- I have variable of overall satisfaction (measured by 5 point rating scale) withing 20 stores
- The overall satisfaction was measured in 4 quarters of the year 2020
- The task is to calculate average satisfaction (score) of the two best quarters within each store
- And then compare the means and determine the significant difference

How should I do this, more precisely which statistical test (probably t-test) should be used?

Many thanks for help!!
 

obh

Well-Known Member
#2
Hi Anetau,

So you compare the average of 20 stores of the best quarter and compare it to the average of the same 20 stores on the second-best quarter?

If so, the two averages are dependent, since it is the same store.
And you need to use the paired-t test. (which is the same as one sample t over the differences)
It actually compares the average of the differences to the expected value (usually zero)

The paired-t-test of dependent pairs is better (more powerful) than the two-sample t-test since it reduces the variation of the average due to the fact you compare different stores. (as you compare on each pair the same store)

You may use the following to check your results.
https://www.statskingdom.com/160MeanT2pair.html