Re: [R] multiple imputation

From: Ted Harding <ted.harding_at_wlandres.net>
Date: Thu, 21 Apr 2011 19:45:04 +0100 (BST)

On 20-Apr-11 20:46:53, DOCMAA wrote:
> I have missing values from a few subjects due to instrumentation
> not working. My data set is N=283 data points. For some subjects
> i have 60 data points missing max.
>
> I tried to use Amelia 2 to impute the missing values but i am
> getting a negative number and i am sure this is wrong because
> its biologically implausible to have a negative number for what
> i am measuring.
>
> Does anyone have any suggestions on how to proceed?
> Thanks.
> --

Amelia assumes that the data are multivariate Normal. This intrinsically allows negative values. One possible way to avoid negative values for a variable which must be positive is to use the logarithm of that variable in your analysis, and proceed as though that had the multivariate Normal distribution.

Then you can back-transform after imputation.

However, whether that is advisable, or whether you should adopt some other approach, can depend on many considerations which can only be inferred from background information about the context in which the data were obtained.

I note that you seem to call the variable with missing values "max". If that is, say, the maximum observed value of a variable over a period of time, then it may be more appropriate to treat it as having an "extreme value" type of distribution, and transform it accordingly.

Better targeted advice might be given if you can supply more detail about what is being observed, what is measured and how, etc.

Ted.



E-Mail: (Ted Harding) <ted.harding_at_wlandres.net> Fax-to-email: +44 (0)870 094 0861
Date: 21-Apr-11                                       Time: 19:44:49
------------------------------ XFMail ------------------------------

______________________________________________
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. Received on Thu 21 Apr 2011 - 18:48:33 GMT

Archive maintained by Robert King, hosted by the discipline of statistics at the University of Newcastle, Australia.
Archive generated by hypermail 2.2.0, at Thu 21 Apr 2011 - 18:50:31 GMT.

Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-help. Please read the posting guide before posting to the list.

list of date sections of archive