R help - Dec 2006 - Multiple Imputation / Non Parametric Models / Combining Results

Dear R-Users,

 

The following question is more of general nature than a merely technical
one.  Nevertheless I hope someone get me some answers.

 

I have been using the mice package to perform the multiple imputations. So
far, everything works fine with the standard regressions analysis. 

 

However, I am wondering, if it is theoretically correct to perform
nonparametric models (GAM, spline smoothing etc.) with multiple imputed
datasets. If yes, how can I combine the results in order to show the
uncertainty?

 

In the research field of real estate economics, the problem of missing data
is often ignored respectively unmentioned. However, GAM, spline smoothing
etc. become increasingly popular. In my research, I would like to use
multiple imputed datasets and GAM, but I am unsure how present single
results. 

 

Again I want to apologize that this is a rather theoretical statistical
question than a technical question on R. 

 

Thanks in advance for any hints and advices.

 

Simon

 

 

 

 

 

Simon P. Kempf 

Dipl.-Kfm. MScRE Immobilienökonom (ebs)

Wissenschaftlicher Assistent

 

Büro:

IREBS Immobilienakademie

c/o ebs Immobilienakademie GmbH

Berliner Str. 26a

13507 Berlin

 

Privat:

Dunckerstraße 60

10439 Berlin

 

Mobil: 0176 7002 6687

Email:  <mailto:simon.kempf@web.de> simon.kempf@web.de

 


	[[alternative HTML version deleted]]

At 08:12 08/12/2006, Simon P. Kempf wrote:
>Dear R-Users,
>
>
>
>The following question is more of general nature than a merely technical
>one.  Nevertheless I hope someone get me some answers.
>
>
I am in no sense an expert in this area but since 
it seems that noone else has answered so far; I 
wonder whether the mitools package from CRAN helps?

>I have been using the mice package to perform the multiple imputations. So
>far, everything works fine with the standard regressions analysis.
>
>
>
>However, I am wondering, if it is theoretically correct to perform
>nonparametric models (GAM, spline smoothing etc.) with multiple imputed
>datasets. If yes, how can I combine the results in order to show the
>uncertainty?
>
>
>
>In the research field of real estate economics, the problem of missing data
>is often ignored respectively unmentioned. However, GAM, spline smoothing
>etc. become increasingly popular. In my research, I would like to use
>multiple imputed datasets and GAM, but I am unsure how present single
>results.
>
>
>
>Again I want to apologize that this is a rather theoretical statistical
>question than a technical question on R.
>
>
>
>Thanks in advance for any hints and advices.
>
>
>
>Simon
>
>
>
>
>
>
>
>
>
>
>
>Simon P. Kempf
>
>Dipl.-Kfm. MScRE Immobilien?konom (ebs)
>
>Wissenschaftlicher Assistent
>
>
>
>B?ro:
>
>IREBS Immobilienakademie
>
>c/o ebs Immobilienakademie GmbH
>
>Berliner Str. 26a
>
>13507 Berlin
>
>
>
>Privat:
>
>Dunckerstra?e 60
>
>10439 Berlin
>
>
>
>Mobil: 0176 7002 6687
>
>Email:  <mailto:simon.kempf at web.de> simon.kempf at web.de
>
>
>
>
>         [[alternative HTML version deleted]]
Michael Dewey
http://www.aghmed.fsnet.co.uk