2

I have been carrying a study and due to a human error in the randomisation process, instead of getting equal sized groups (I have 8 and aimed to have 100 per group), I now have around 90-100 participants per group (in 7 groups) and for one group only double that amount (large group - 200 participants). The study looks at different messaging around an illness concept, no patient involvement. I am to do ANOVA's and MANOVA's for this.

A few options I have from here:

  1. do the analysis as I am not violating ANOVA's assumptions (I should be fine)
  2. cut participants from the data in the 'large group'. I would exclude people who were recruited by the platform we were using based on time, thus the last 100 to be removed (Unsure)
  3. recruit more people for the other 7 groups (I would avoid due to costs but can still be an option)

What would be the best option? Further, for the 2nd there are some issues but cannot find appropriate literature to consult. What are the main issues with this? Apart from intention-to-treat principle and having an ethical obligation to participants since they were told that their involvement will help scientific research, I have not found much else.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Cristina
  • 21
  • 1
  • 1
    What kind of error exactly? Is it just that one of the groups is larger, but the assignment was still random, or maybe it was non random, or even invalid? – Tim Sep 08 '21 at 08:13
  • Assignment is still random and representattive, I looked at the data. The mistake was that I thought I set the quota in ths study (but I did not) so that each group gets the allocated 100 people. One group larger indeed, and assignment still random – Cristina Sep 08 '21 at 09:08

1 Answers1

2

Equal group size is not a requirement for anova (it might be an advantage if you can have it). So, as long as group assignment is still random and representative, as you say it is, there is no reason to do anything.

Just continue to do the analysis.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467