www.cmm.bristol.ac.uk/forum

I found it very interesting and helpful as I read about rules-of-thumb to determine the appropriate sample size for conducting a multilevel analysis on MLWin. The link I refer to: http://www.bristol.ac.uk/cmm/learning/m ... mples.html

These hints, however, only relate to level 2 units and motivated me to ask how you would evaluate my current sample size. In particular, I have unbalanced panel-data: 128 individuals (level 2) and 326 observations (level 1) which are measured at three waves. This indicates that I have 2.5 observations per individual. Does this - in your opinion and based on your experience - relate to a reasonable statistical power of e.g. 0.8?

My goal is related to getting non-zero results on the basis of a within-between random effects specification with Bayesian analysis. Furthermore, I plan to conduct my first-time retrospective power analysis via MLpowsim.

Best,

Sabine

Hi Sabine,
That is really hard to say without knowing what you are hoping to test. Statistical power links to a test and will depend on what that test is and what effect size you are looking for / degree of clustering etc. There should be details in MLPowSim that will help.
Best wishes,
Bill.