Fisher’s ex

[sumah]

Any clue when is Fisher’s exact test invalid or in other words, when it can NOT be used? I’m getting a warning message about % of missing data in the SAS output but couldn’t figure it out! I used proc freq

[pink_poodle]

Could you please write what it says in the warning message?

[sumah]

[sumah]

[sbxkoenk]

Hello,

Can you present us with the exact warning message?

Is it a WARNING in the LOG or in the OUPUT / RESULTS??

I understand SAS is complaining about missing values in your data set, not about frequency counts that are zero nor about other sparsity in the data, correct?

Thanks,

Koen

[sumah]

[sbxkoenk]

Hello,

That "warning" in the results is just to make you aware that not all your observations were used in the PROC FREQ Fisher's exact test as not all your observations can be counted (some observations have missing values for the row and / or column variable in your contingency table).

It does not mean that the Fisher's exact test is invalid.

It is valid but not all observations participated in the analysis!

Koen

[sumah]

I see! I usually use Fisher's test and that was the first time I get a warning and in the output, not the log. Thank you so much!!! 

[Ksharp]

That said ChisquareTest is not valid, NOT Fisher Exact Test !
ChisquareTest and Fisher Exact Test is totally different !

[sbxkoenk]

Thanks for pointing that out @Ksharp .
indeed, an important addition on your part!

Koen

[sumah]

I started with ch-squared and it wasn't valid so I switched to Fisher's test

[Ksharp]

But you said:
"Any clue when is Fisher’s exact test invalid "

You can always use Fisher Test as long as it is displayed .

Chisq Test require 5% count in each cell as least .

[sumah]

yes, that was my question, I thought at first I can't run a fisher's exact test but found out it's okay. thanks again.