Dear Robin,
You already have a literal answer to your question, which is to look at
MIcombine.default, but this is just implements Rubin's rules for combining
multiple imputations, which are described in most treatments of the subject.
What's curious is that with only one missing observation among 1409, you
would have a fraction of missing information for the coefficient of u of
64%. Of course, the fraction of missing information for a coefficient isn't
simply the fraction of observations missing for the corresponding variable,
since the coefficient for that variable will be affected by missing data on
other variables, but if the several multiple imputations produce very
similar coefficients, as should be the case when only one observation is
missing, then the fraction of missing information should be small for all
coefficients.
So something is out of whack here. In particular, you don't say how you got
the multiple imputations in rt.imp, and to what extent the coefficients
produced from them differ. I suspect that either some mistake was made, for
example in preparing the data, or that the data must be extremely unusual in
some respect. If I were you, I'd start by looking at how the results for the
completed data sets in rt.imp differ (e.g., the coefficients of u must
differ a lot), and if that doesn't reveal the problem, at the completed data
sets themselves.
I hope this helps,
John
--------------------------------
John Fox
Senator William McMaster
Professor of Social Statistics
Department of Sociology
McMaster University
Hamilton, Ontario, Canada
web: socserv.mcmaster.ca/jfox
> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at
r-project.org]
On> Behalf Of Robin Jeffries
> Sent: November-07-10 1:24 PM
> To: r-help at r-project.org
> Subject: [R] How is MissInfo calculated? (mitools)
>
> What does missInfo compute and how is it computed?
> There is only 1 observation missing the ethnic3 variable. There is no
other> missing data.
> N=1409
>
> > summary(MIcombine(mod1))
>
> Multiple imputation results:
> with(rt.imp, glm(G1 ~ stdage + female + as.factor(ethnic3) + u,
> family = binomial()))
>
> MIcombine.default(mod1)
> results se
> (lower upper) missInfo
> (Intercept) -0.40895453 0.14743928 -0.70805544 -0.1098536
> 53 %
> stdage 0.13991360 0.06046537 0.02140364
> 0.2584236 0 %
> female -0.05587635 0.11083362 -0.27310639
> 0.1613537 0 %
> as.factor(ethnic3)1 0.17297835 0.19556664 -0.21032531 0.5562820
0> %
> as.factor(ethnic3)2 0.63507020 0.18017975 0.28192410 0.9882163
0> %
> u -0.01322976 0.18896230 -0.40291914
> 0.3764596 64 %
>
> Thanks,
>
>
> Robin Jeffries
> MS, DrPH Candidate
> Department of Biostatistics
> UCLA
> 530-624-0428
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code.