Chi Square/ Mcnemars or Cochran's Q test for dependent groups

machine_learning

#1

Hello All,

I have this scenario.

Year       Num_Insured        Num_claims        Proportion

2015             50                20              0.4

2016            120                40              0.33

2017            140                110             0.79

I want to test whether the proportion of num of claims to num of insured (proportion column) remains the same throughout the years

There could be common customers from one year to the next so I assume that makes this a dependent groups so I cannot conduct a regular chi square. I am thinking if Cochran’s Q test will work here; am I assuming correctly that mcnemars wont here as it is 3 or more years (3 or more groups?)). Any ideas would be highly appreciated…

Regards


#2

If I understand correctly, you want to check the values in the ‘Proportion’ column are the same for all rows or not. is that correct?


#3

Yes, that would be correct. May be I should do each test for two years at a time, that compares proportions from one year to the next. So there will be 3 tests i.e. from 2015 compared to 2016, 2016 to 2017…but which test would it be…