Hi Jennifer, I'm not entirely sure of my answer but I recall reading somewhere, that you have two options; mean substitution for those with missing data ie calculate mean of the other respondents and this value is inputted for those with missing data. Otherwise just do the analysis for the 524 at both time points and delete the rest. I use Armitage and Berry; statistical methods in medical research as a reference but I've not gotten my answer for this from there. Do you have it in your library? I have it, but it is falling apart as the binding glue disintegrated.

Hello Jennifer,

If the respondents were anonymous, then the best you can do is treat the two batches (2022, 2023) as if they were independent of one another, and run your comparisons by considering them as distinct groups.

Of course, it's quite possible that the two samples are not equally representative of the target population (due to factors such as self-selection), even if you had come up with approximately equal numbers of respondents each year.

In an ideal world, you'd be able to match those folks who responded both times and rely most heavily on their results (for drawing inferences about change in perception). However, as most surveys are anonymous, this isn't feasible for most studies. Exceptions would be longitudinal studies in which the same participants are followed up over time.

Good luck with your work.

Jennifer Bateman wrote: I'd like to compare to see, for example which job roles changed from the highly valuable answer choice to the valuable answer choice.

This raises several questions (in my mind, at least):

• What is the dependent variable?
• How many job roles are there?
• What type of analysis would you use if you had had two data points per person with an ID code allowing you to do a paired analysis?

Thanks for clarifying.

Deborah J Hilton, I think that all experts on missing data eschew mean substitution these days. See John Graham's nice Annual Review chapter, for example.

Article Missing Data Analysis: Making It Work in the Real World

OK I will read this Mr/Dr Bruce Weaver when I can. I have vertigo today so it could be a bit much to read just now, maybe I'll read new idea now, that later

I've also been plugging away at trying to get the local church group to be more evidence based/ scientific & have to prepare something for them to read, although when I've done this previously, no one reads it properly.