0

I read Non-parametric test if two samples are drawn from the same distribution

and have tried permutation test, KS test

The issue is I have data of different sample size that are collected from the same subject, hence dependent.

Eg.

Condition A: 180 samples

Condition B'0: 23 samples, Condition B'1: 30 samples Condition B'2: 25 samples

Condition C'0: 23 samples, Condition C'1: 30 samples Condition C'2: 25 samples ...

So the two variables in this case is

  • Letter [B,C,...M]
  • Number [0,1,2...N]

Scenario 1

For condition B*, I can assume they can be pooled, creating Condition B of 78 samples.

Now I want to test if condition A and pooled condition B are from the same distribution.

Scenario 2

I want to pool \$letter'1 together, and \$letter'2...N together and compare between the pooled condition. For example, [B'1, C'1...,M'1] Vs. [B'2,C'2...,M'2] etc.

My understanding is that KS test is for independent samples (including Two-sample Kolmogorov–Smirnov test).

I was wondering if there are any tests that can test this on dependent samples with unequal size

Elder Gu
  • 1
  • 1
    What kind of dependence occurs here? Being from the same subject does not make them dependent. It only means you cannot employ the standard justifications to make inferences to any other subjects. – whuber Nov 27 '22 at 17:33
  • there are multiple conditions. Condition A is the resting state. Condition B'1/B'2/B'3 are task state recorded for same subject, Condition C'1/C'2/C'3 are recorded on the same subject. I think to your point, resting state maybe considered independent from task conditions, but say I want to pool B'1,C'1...M'1 together and compare with B'2,C'2...M.2, that would be dependent right? – Elder Gu Nov 27 '22 at 19:12

0 Answers0