Page 1 of 1
Fisher's exact test not running
Posted: Sun May 21, 2023 12:40 pm
by Victor2
Let me start by thanking the creators for providing jamovi!
I am trying to run a Fisher's exact test on a fairly large dataset (with many expected frequencies <5). However, the test does not appear to run (a P value is not appearing). When I remove some data (e.g., entries with frequencies >10), the test runs properly. I have tried to let my (fast) computer run for a long time, but still no P value is appearing. Is there perhaps a maximum dataset size for this test in jamovi?
Re: Fisher's exact test not running
Posted: Sun May 21, 2023 1:54 pm
by JorgeCamachoS
Hi
Fisher Exact Test apply only to small tables, in jamovi only for 2 x 2 tables. If you are interested in species x sex, Fisher test is not appropriate
Re: Fisher's exact test not running
Posted: Wed May 24, 2023 1:37 pm
by Victor2
Thank you for your reply. However, jamovi has no problems performing Fisher's exact tests on larger tables than 2 x 2 and the test appears appropriate for my data. I just do not understand why it does not run in the example that I gave, whereas it does run when I remove a small number of rows.
Re: Fisher's exact test not running
Posted: Wed May 24, 2023 4:10 pm
by MAgojam
Victor2 wrote: ↑Wed May 24, 2023 1:37 pm
However, jamovi has no problems performing Fisher's exact tests on larger tables than 2 x 2 and the test appears appropriate for my data...
Hey @Victor2,
what you say is right and generally there is final p-value in the table.
The analysis of your data does not end as expected with a p-value, because the
default work area is insufficient for the necessary calculation process (unfortunately, there is no note in the table regarding this lack of p-value ).
I can tell you that I'm talking to @Jonathon about a
code fix based on a possible workspace increase or (what I prefer) switching to a Monte Carlo simulation, when this error occurs, to arrive at a p-value final in table.
In the screenshot the result of a Monte Carlo simulation with data that could be yours.
- ScreenShot_20230523104030.jpeg (120.01 KiB) Viewed 1267 times
Thanks for reporting this to us.
Cheers,
Maurizio
Re: Fisher's exact test not running
Posted: Wed May 24, 2023 4:47 pm
by JorgeCamachoS
Re: Fisher's exact test not running
Posted: Wed May 24, 2023 5:21 pm
by Victor2
Many thanks for solving the mystery @Maurizio, much appreciated! I will be looking forward to the code fix.
Re: Fisher's exact test not running
Posted: Thu May 25, 2023 9:37 pm
by Victor2
MAgojam, perhaps a stupid question, but how have you been able to do the Monte Carlo simulation in jamovi? When I used the Rj module for the simulation, I end up with a "FEXACT error 40. Out of workspace".