In SPSS never gives an automatically Fisher exact test, when your data have 20% of less than 5 counts. You should give the option to calculate Fisher exact test. When you open your "crosstabs" option in descriptive analysis, you can see the "Exact" clicking button on above the "Statistics" option. You should open the "Exact" and click one of the tests of "Exact" . Now you can select asusal chi-square test in "statistics" and run the test. You can get Fisher exact test results also in the output table.
In SPSS never gives an automatically Fisher exact test, when your data have 20% of less than 5 counts. You should give the option to calculate Fisher exact test. When you open your "crosstabs" option in descriptive analysis, you can see the "Exact" clicking button on above the "Statistics" option. You should open the "Exact" and click one of the tests of "Exact" . Now you can select asusal chi-square test in "statistics" and run the test. You can get Fisher exact test results also in the output table.
The only problem with applying Fisher's exact test to tables larger than 2x2 is that the calculations become much more difficult to do. The 2x2 version is the only one which is even feasible by hand.
Nevertheless, the test can be applied to any mxn table and some software including Stata and SPSS provide the facility. Even so, the calculation is often approximated using a Monte Carlo approach.
When you have many cells or many units then the Exact becomes time consuming to compute (as in a computer takes a long time to do it), as you have to list every possible outcome. I know I did it many years ago for my MSc project . With Monte Carlo you just take a sample of the possible outcomes rather than all and base your p-value on that.
Well you do not have to worry about observed count for any contingency table, if expected count are less than five spss gives fischer exact test for interpretation of your p value.
I tried this but did not get the Fisher's Exact Test as an option in my Chi-Square Tests...only got Pearson Chi-Square, Likelihood Ratio, and Linear-by-Linear Association. What am I doing wrong?
Dear Denise Lewis, when you go in crosstab option, then in new pop up window you have to click first on "exact" option then again click on exact option to get p-value of fisher exact test.
Basically for 2x2 table SPSS provide automatically Fisher's exact test whereas for 2x3 and above, we have to click on exact test option in crosstab box.
when in RxC table, if any cell have expected count less than 5 then Chi square test does not valid. in this situation you have to use Fisher's exact test that is valid test in this case.
Go to analyze then Crosstab then shift variable into row and column box , then click in chi square and exact test and then press ok.
After a Fishers exact test has been carried out for a 6 x 2 table and given a significant value, is there a way of finding out which aspect has caused it to be significant?
Yes. you can find out by creating 2x2 table for each factor. after creating 2x2 tables for each factor apply chi square or fisher exact test. but this time you have to use Bonferroni correction to adjust the p-value for each test.
Hi Waqas, thank you for your reply. I am using SPSS and have created a 2x2 table for each factor (so 15 for the overall 6x2) and need to use Fishers because of low sample size. The output in SPSS for the 2 sided significance gives 1.000 which I do not understand? How do I apply Bonferroni to this. Your help is much appreciated
P-value 1.000 means that there is no association between both variables.
for association it should be less than from level of significance that is usually 5% (i.e., 0.05)
For Bonferoni correction, we multiply the significance value of output in SPSS by number of your 2x2 tables. in your situation every significance value should be reported after multiply by 15. if after multiplication it exceed from 1 than you just report p-value > 0.999 and conclude that there is no association between both variables.
Ma'am. You can leave it as a blank. That is easy when you will be doing analysis. Because you don't want to give any condition for doing statistical analysis without missing values.
I'm not sure why no one mentioned it before, but SPSS has a plugin you can download. See here: http://www-01.ibm.com/support/docview.wss?uid=swg21479647
Fisher 2x2 was traditionally the only option, but modern computers crunch larger sizes no problem. And SPSS DOES provide Fisher's when there are too few numbers in a cell, but only for 2x2. That's the whole point of Fisher's exact.
and as for missing data, you can code it as anything you like, personally I use 999, any value that's far outside any expect values, but then you have to assign that number as "missing data" number in the variable view when you're labeling other data. If you don't tell SPSS that it's a missing data number, it will include that number like an actual data point.