[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: Re: st: how to calculate marginal effectsfor the multiply imputed survey dataset

From   <>
Subject   Re: Re: st: how to calculate marginal effectsfor the multiply imputed survey dataset
Date   Mon, 19 May 2008 11:15:36 -0500 (CDT)

Hello Prof. Wagner,
Thanks for sharing your experience. I might also consider  working with odds ratio.




What about using svy logistic and presenting odds ratios?  In my 
field (health), most people use odds ratios and find them easy to 
discuss with dummy covariates. 


At 02:48 PM 5/15/2008, you wrote: 
>I am using 2003 Survey of Small Business Finances data. This is a 
>survey data with multiply imputed dataset for 5 imputations, 
>stratums and weights. Furthermore, I am using Stata 9 version. Below 
>is my question: 
>I am running a survey logit and survey probit estimation for this 
>multiply imputed survey data.  However, for the overall estimation 
>for the multiple imputation, I cannot get the marginal effects 
>estimation. I can only get this for the 5 imputations separately in 
>Stata. However, there is no way I can get the marginal effects for 
>the overall estimation. I even looked at this in SAS. For instance, 
>SAS doesn't give me at all any marginal effects, even for the 
>separate implicates (I used the proc surveylogit and proc 
>surveyprobit commands for each separate implicate and then combined 
>these results through mianalyze command in SAS for the overall estimation). 
>However, I was able to get the marginal effects for separate 
>implicates in Stata (I first used svyset command, then miset and 
>then mim commands for the overall estimation. I even tried using 
>mifit command rather than mim command. However, I couldn't find any 
>postestimation command for marginal effects). 
>Do you know whether in literature, marginal effects have been 
>calculated for the overall estimation for a multiply imputed 
>dataset? If yes, can you pls lead me about how to do the marginal 
>effects estimation for the overall estimation for my survey data 
>with multiple imputation either in SAS or STATA?  I will really 
>appreciate any type of help I can get. 
>Thanks a lot. 
*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index