Hi all,
I have conducted a -teffects ipw- analysis on an observational dataset. I am using -tebalance summarize- to obtain standardized differences, which make sense to me (and for which I can replicate the calculations) for both continuous and dichotomous categorical variables.
However, for multi-level categorical variables, I would like to present a single standardized difference. The -tebalance summarize- output, however, provides a standardized difference for each non-base level of the categorical variable. See below:
In the example above, "LV_grade" contains four levels.
I found the following PDF which seems to address my issue, but I am uncertain how to extend the special case of a two-level categorical variable to a multi-level variable. http://support.sas.com/resources/pap...2/335-2012.pdf
Is anyone able to help?
Many thanks, Phil
I have conducted a -teffects ipw- analysis on an observational dataset. I am using -tebalance summarize- to obtain standardized differences, which make sense to me (and for which I can replicate the calculations) for both continuous and dichotomous categorical variables.
However, for multi-level categorical variables, I would like to present a single standardized difference. The -tebalance summarize- output, however, provides a standardized difference for each non-base level of the categorical variable. See below:
Code:
-----------------------------------------------------------------
|Standardized differences Variance ratio
| Raw Weighted Raw Weighted
----------------+------------------------------------------------
|
LV_grade |
2 | -.1980125 .1603124 .8048659 1.128188
3 | -.165836 -.1739039 .60628 .5127054
4 | -.0760321 -.0712021 .4732641 .3883542
I found the following PDF which seems to address my issue, but I am uncertain how to extend the special case of a two-level categorical variable to a multi-level variable. http://support.sas.com/resources/pap...2/335-2012.pdf
Is anyone able to help?
Many thanks, Phil
