If you are familiar with the Multiple Imputation Tool in the Missing Value
Analysis procedure of SPSS, I need your advice or recommendation. Whereas
the tool provides pooled results of the multiple imputed data sets for
some statistics of the multiple imputation models (e.g., unstandardized
coefficients, signicance of the coefficients, partial correlations are
provided for multiple regression) it does not provide some other important
statitics in pooled form (e.g., Standardized co-efficients or Betas,
Collinearity measures such as VIF or collinearity tolerance). What is a
bit confusing is that these unpooled statistics are provided itemized for
each imputation model. If I simply averaged these stats manually across
the imputation models for a given variable (since the pooled statistics
SPSS provides is also an average accross the imputation models of a given
variable, would that be statistically acceptable? Is there a reason why
SPSS pools the unstandardized coefficient but not the stardandized ones
(Betas)?
Thank you so much in advance for your help!
George
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
