I have some categorical data from a clinic taken over a year long time period. The plan is to perform a Pearson's Chi Squared test (of independence) on this data.

I am extracting data from a clinic's database and my aim is to see if male patients are more likely to cancel their appointment than female patients.

My query is this: Chi squared requires independent samples but there is a chance that patients may have had multiple appointments within that time period. Whilst I could identify these multi-attenders, how would I then assign them to any one category (presuming some would appear in multiple different categories ).

For instance...

John Doe may have:

  • Attended a cardiology appointment in January
  • Attended a cardiology appointment in February
  • Cancelled a cardiology appointment in March

To push this to it's limit:

  • John Doe then goes through a (very fast) gender reassignment procedure and becomes Jane Doe and then cancels their cardiology appointment in April.

So from one person we have:

  • Male/Cancelled: 1
  • Male/Attended: 2
  • Female/Cancelled: 1

Is Chi Squared the correct test to use in this instance? If so, is there some data processing I should be performing on my data to account for patients who appear multiple times in different categories?


1 Answer 1


A statistical solution to the (main) problem of multiple appointments is to look at proportion of appointments cancelled. However, if most people have only one appointment, then this will have huge lumps at 0 and 1. A possible solution to that is to use beta regression, which deals well with data that are bounded.

A data structure solution to the main problem is to use, e.g. canceling First appointment as the variable; or possibly cancelling any appointments, or something else, depending on the goal of your research.

If you really do have gender reassignment then there are at least two possibilities: 1) Do not divide "sex" into just "male" and "female" (there's lots of literature on this type of topic, if you really need it) or 2) Just delete these people from the table and write about them in the text.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.