Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: Multiple imputation int() option in ICE
From
Nick Cox <[email protected]>
To
"'[email protected]'" <[email protected]>
Subject
RE: st: Multiple imputation int() option in ICE
Date
Thu, 18 Nov 2010 11:47:29 +0000
It sounds to me as if there are several issues conflated here.
It comes as a surprise to hear that BMI is normally distributed. Is it really true?
That said, are you sure that you are not just seeing some arbitrary codes here, such as 0 for "very low but not measured" and 1000 for "very high but not measured". True or not, such values sound like values that should be recoded to missing, and not bounds to be respected in imputation.
After all, we all have folk stories along these lines. I read of a story in which average age of deaths in traffic accidents was alarmingly high until it was realised that 99 was being used as a code for "not recorded" but being averaged in nevertheless. Naturally, Stata users would know better.
Nick
[email protected]
Maarten buis
--- On Thu, 18/11/10, Alberto Osella wrote:
> I wish to impute a data set wit missing data in several
> variables of diferente type; one of them (BMI) is a
> continuos variables with a normal distribution. Obviuosly I
> should to bound BMI because there are some impossible values
> as 0 or 1000.
Strictly speaking, the presence of these bounds means that the
distribution is not normally distributed. An alternative way of
ensuring the bounds is to specify the -match- option. I think I
remembered that there was quite a discussion on it in the Stata
Journal articles that accompanied ice, see:
<www.stata-journal.com>
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/