Fisher's exact test is utilized when there is a need for a chi-square test, but one or more than one row in your observation dataset have five or less values in terms of frequency. The basic assumption in a chi-square test is that the frequency of the values in the rows of the given dataset is five or more than five. Fisher's exact test does not need this assumption:

tabulate model fridge_type,
exact

Here are the results of this query:

This outcome tells you that fridge_type and model are not statistically related (that is p is equal to 0.652). Fisher's test computes the p-value directly. Here is the mathematical derivation of the fisher's test:

Look at the following assumed data of fridge sales:

Then, the probability and the p value is given by the following formula: