Hi all,
Does anyone have any suggestions as to how to analyse a data set with missing values, where the observations are correlated and the correlation structure is known? My feeling is that it would be interesting to use multiple imputation but to develop an imputation method that explicitly incorporates a correlated error structure, using generalized least squares or (more generally and bravely) GEE methods. However, before I begin the long process of trying to translate my ideas into computer code, I would like to know whether I would be reinventing the wheel (or inventing the exactly wrong wheel!).
Any comments are welcome.
Thanks,
Simon.
Simon Blomberg, PhD
Depression & Anxiety Consumer Research Unit
Centre for Mental Health Research
Australian National University
http://www.anu.edu.au/cmhr/
[log in to unmask] +61 (2) 6125 3379
|