Missing Data A: I want you to use all the data you can use for this analysis. If people are missing some values can your statistical modeling handle it?
A: Instructor answer - no imputation required for this project. It will make the project take far too long in terms of hours. If you think imputation will improve the analysis please put your rational in the discussion of your report.
Are there non-linear effect?
A: It isn’t clear to me, the investigator, that the treatment effect is the same over time. I think that means it could be non-linear. The picture I have in my head is that the treatment works well early and then works less well later. That is just a guess though. Can you tell me the pattern of the treatment effect over time?
How do I measure responsiveness to a bronchodialator?
A: Responsiveness to a bronchodialator is measured by the difference in the post - pre measures of FEV1?
How do you want me to use time in my statistical models (categorical, continuous)?
A: Continuous in my final report and measures can be reported as change per year or change per month, whatever makes the most sense based on the time scale.
Other covariates
A: Same as for Project 1.
When does treatment end?
A: Treatment cut-off is the same as for Project 1. Values after this cut off are follow-up observations.
Handling drop out.
A: Please do not impute the last observation as the values for future missing events (called last observation carried forward).
Time periods for the study
A: Conceptually there are two time periods for this study. The treatment time period and the non-treatment follow-up time period. My first hypothesis is about the treatment time period. My second hypothesis is about the 2nd time period.
What are the secondary outcomes?
A: FEV1/FVC, post-pre bronchodialator
Who should be in the analysis?
A: Everyone. We realize that for the follow-up part that some people have no follow-up data. Please make me to Table 1 tables for each part of the project.
What should we do with the survey data which is intermittently collected?
A: Please create one variable for each to use across time. If you want to use all available measures come to see Tasha or Nichole to discuss.
A: You can ignore it for analysis