0
$\begingroup$

I'm looking to compare the read rate of an email newsletter with the read rate of another newsletter. However, many (not all) of the people who receive one newsletter also receive the other. What would be the best method to determine whether the proportions of people who read the newsletters are significantly different, given that there is a large overlap in the samples?

I know of McNemar's test, but that requires matched pairs of subjects. The newsletters' datasets are stored differently, so I unfortunately can't see whether an individual who has read one newsletter has also read the other.

Thanks!

$\endgroup$
0
$\begingroup$

This is a bad situation. If the values are missing because some people do not read one of the newsletter then there is nothing to do here, you couldn't impute these values for ex.. Dropping these people (that do not read both newspapers) could be an option only if they represent a small proportion of your data (5 to 10 %). Otherwise you may have to use a non-paired test, where this would not be a problem.

$\endgroup$
  • $\begingroup$ I agree, it is a bad situation. The dataset for one of the newsletters is very bare-bones. It shows whether a user received the newsletter and whether they opened/read it, but it doesn't give any information about that user other than an ID, which is different from the IDs in the dataset for the other newsletter, so I cannot pair the datasets by user. Is there a specific non-paired test that you think would work best? Thanks! $\endgroup$ – MattStan Jan 30 at 15:29
  • $\begingroup$ @MattStan A standard t-test would be the first choice, if the assumptions are not met then a non-parametric version such as the Mann-Whitney test. However, keep in mind that non-paired tests assume independence between samples, which is not true in your case (for some proportion of your data which read both newsletters), so take the results with a grain of salt. $\endgroup$ – user2974951 Jan 30 at 15:34
  • $\begingroup$ @user2974591 Thank you so much for your time and help! I greatly appreciate it! $\endgroup$ – MattStan Jan 30 at 15:45
  • $\begingroup$ @MattStan If this answer helped you in solving your problem consider accepting it by clicking the tick button on the left. $\endgroup$ – user2974951 Jan 30 at 15:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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