this study aim to evaluate a new intervention for back pain where clinical outcome measures such as disability, pain and eq5d (all of these are continuous data) were collected at baseline, three and six months later
the preliminary results showed that patients with higher pain and disability at baseline did not complete the study.
we started with 32 and end up with 20, what is the best method for missing data management knowing that this study is not powered ti detect difference in outcomes
Hi
thank you for your replay
I was trying to do MI in stata, however I was confused about which option to choose in the impute option, I'm not planning to do any regression because of the small sample size.
I just want to impute the missing data to fully describe my clinical outcomes
This is a small sample alright but it does not mean that data cannot be analysed quantitatively. Publishing these findings can be difficult unless this study gets referred to as a feasibility or pilot study (especially that you refer to it as a new study)
I hope this helps
M.
Dear Mohammad
thank you for your replay, the data will be analyzed quantitatively however, we can do any inferential statistics because study is not powered to detect any difference in clinical outcomes and the protocol of this study is already published as feasibility study
best regards,
The data is not missing at random, so imputation might be probelmatic. I would think of a model that accounts for compliance. I have no idea if there are some "ready-made" models for this purpose, but as this problem is not so rare I would guess that there are such models.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis