Dear Dionne: there is some inconsistency in the literature on this point. Rubin (who originally defined this terminology) would say (2) is MCAR and (1) is MAR, but others have defined (1) as MCAR. Rod Little On Thu, 2 Aug 2001 [EMAIL PROTECTED] wrote: > > > I have been discussing the definitions of MAR and MCAR with others. Data > missing by design is an example of MCAR that is given often in the literure. > Some think this is not an example of MCAR, but an example of MAR. > > >From my understanding of the definition of MCAR, data is MCAR if > > (1) P(R| Y, Z) = P(R |Z) > > where Y = complete data matrix, R = missing data indicator matrix for Y, Z = > matrix of conditioning covariates,experimental treatments, and other design > variates. Hence, the above example is MCAR. > > However, my colleagues think that (1) is the definition of MAR. and the > definition of MCAR is > > (2) P(R|Y,Z) = P(R). > > In this case, the above example is not MAR. > > > In the presence of conditioning covariates, and design variates, what is the > definition of MCAR? > > > Thanks, > Dionne > > > > > > > > > > > ___________________________________________________________________________________ Roderick Little Chair, Department of Biostatistics (734) 936-1003 U-M School of Public Health Fax: (734) 763-2215 M4208 SPH II [EMAIL PROTECTED] 1420 Washington Hgts http://www.sph.umich.edu/~rlittle/ Ann Arbor, MI 48109-2029
