Dealing with missing data in Repeated-measures ANOVAs


New Member
Afternoon all,

I've been running a longitudinal follow-up study to assess outcomes following a psychological treatment for ME. As to be expected, there was a lot of drop-out by 12 months, the 5th time point of the follow-up. Usually I'd do a RM ANOVA to analyse this, but with the missing data and drop-out, it leaves my sample very small by the end, whereas it's fairly robust earlier on at earlier time points with a much better sample size. What can I do or should I be doing another test?



New Member
ME is myalgic encephalopathy - often allied with chronic fatigue syndrome.

I started with about 140 participants, but think I'm down to about 60 at 12 months (can't check exactly as I'm at home for the day now!). A heavy drop out, but not uncommon in clinical work unfortunately


Super Moderator
There is quite a lot of literature concerned with the problem of analysing data from a clinical trial where dropouts are present. You have a few options, and probably need to do some reading about them to decide what's best in your case. Certainly you shouldn't just be analysing the complete cases.

Some options include:
Intention to Treat analysis
Last Observation Carried Forward
Mixed Model for Repeated Measurements
Multiple Imputation

Do you have a control group?