SPSS insufficient memory Fisher exact test

[brett_a_baxter]

Hi All,

I have been conducted analyses on a study using the Fisher Exact test but have currently encountered an issue with my larger contingency tables (3x7 and 3x9) as SPSS has insufficient memory to conduct the analyses. In this scenario it is best to use a Monte Carlo simulation on these contingency tables and if so, should I now go back and use Monte Carlo simulation on the previously analyses contingency tables that the Fisher exact test was used for?

Thank you for your help in advance,

Kind Regards,

brett

 

[fed2]

try getting a computer with more memory.

 

[hlsmith]

They also didn't mention sample size. Switch over to chi-sq test. Factorials are tearing your computer apart.

 

[fed2]

probably mc method is the way to go. you can mix and match monte carlo / exact.

 

[brett_a_baxter]

They also didn't mention sample size. Switch over to chi-sq test. Factorials are tearing your computer apart.

There are greater than 20% of responses with fewer than 5 responses so I have avoided chi-squared test in this instance.

The sample size is 405.

 

[brett_a_baxter]

probably mc method is the way to go. you can mix and match monte carlo / exact.

Thank you for your advice.

 

[fed2]

Pro tip: Set the random seed for MC method, so you can recreate later.

 

[hlsmith]

You are probably fine, but it is < expected 5 value not empirical counts < 5. And that general rule has been questioned more recently like other rules of thumbs created before big simulation studies.

IBM has a quantum computer people can sign up to use. I kind of wanted to try and code out Fisher's test and try to run it on that computer to see how fast it might be using qubits.

 

[Karabiner]

There are greater than 20% of responses with fewer than 5 responses

In more than 20% of CELLS there are fewer than 5 EXPECTED frequencies,
I suppose SPSS was telling you.