I am attempting to test for a significant pattern in a reduction of prescribing drug A and an increase in the prescription of Drug B.

I have proportions of patients started on drug A vs. drug B vs. neither for many consecutive years. Time is my independent variable and the proportion is my dependent variable. Is there a way to test for statistical significance?

Example of my data:

1996: 0.4 drug A, 0.3 drug B, 0.3 none

1997: 0.3 drug A, 0.4 drug B, 0.3 none

1998: 0.2 drug A, 0.6 drug B, 0.2 none

1999: 0.1 drug A, 0.8 drug B, 0.1 none

  • $\begingroup$ "Measuring" is not a testing problem but an estimation problem. $\endgroup$
    – Glen_b
    Jan 24, 2017 at 1:36

1 Answer 1


If you want to describe the pattern, you might simply plot the data. If you have the N's that go into the proportions, you could do a 3 by number of years chi sq to see if the variations over years is greater than expected based on chance. The Chi Sq test assumes independence of observations, so you would need to think about whether your data meets or violates that assumption. For example, do any of the patients contribute data for more than one year?

  • $\begingroup$ Thank you. I have one more question. How would I determine the expected variation based on chance? I am testing to see any difference between the years so would the expected variation be 0? $\endgroup$
    – vrm
    Jan 24, 2017 at 21:06
  • $\begingroup$ The probability associated with the Chi Sq value is typically used to help decide if your observed variation is reasonably thought to be due to chance. $\endgroup$
    – Joel W.
    Jan 25, 2017 at 14:19

Your Answer

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

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