Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: Missingness


From   Zachary Levy <LevyZ@wrapsnet.org>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   RE: st: Missingness
Date   Tue, 28 Aug 2012 13:15:32 +0000

As Nick suggested, if you have a significant number of response variables missing and your explanatory variables are not, then you can develop a model to look for some pattern to the missingness. 

The reason you might do this is to drop the observations with missing variables more comfortably with the claim that they are missing at random, which makes your results more defensible. If your findings show that there is a significant relationship between your explanatory variables and the probability of being missing, then your regression results will be biased for the ordered logit (or some other approach) model that you are specifying with the Likert scale.

Zach

My views do not neccesarily reflect the views of my employer.
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2..harvard.edu] On Behalf Of Nick Cox
Sent: Tuesday, August 28, 2012 4:56 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Missingness

You have therefore observations with missing values on your response or outcome variable. It doesn't matter what extra predictors you cook up, so far as I can see, as you can't include such observations in your main model. That doesn't rule out some comparison of people included in the model and people not included in the model. One way to do that is an extra model, say logit of (missing or not on outcome) on predictors in model, which may be what you have been reading about.

Nick

On Tue, Aug 28, 2012 at 9:16 AM, Brendan Churchill <Brendan.Churchill@utas.edu.au> wrote:
> sorry I wasn't very clear
>
> I have ordinal measures on a Likert scale 1-7 as my dependent variables, which I am treating as interval for the purposes of analysis. There are some negative values on these measures which indicate that a respondent either gave more than one answer on the measure or did not complete this section of the survey. I usually recode these as missing =.
>
> I don't think I should treat them as zero because wouldn't' that affect my interpretation of the dependent variable?
>
> Thanks for the response
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nick Cox
> Sent: Tuesday, 28 August 2012 6:07 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: Missingness
>
> Your strategy isn't clear. Regardless of whether or how you use an extra missingness variable, how do you expect Stata to treat the missing values in the variables you already have?  Also, are the ordinal predictors being treated as ordinal? Is the response ordinal and does it include missing values too? One way forward is to treat "missing" as just another category with its own code, but that would seem to oblige you to treat such variables as nominal (in practice as equivalent indicator variables) -- unless somehow you know that "missing" always means "very big" or "very small" or "zero" and so can be placed or one end or within the order.
>
> Nick
>
> On Tue, Aug 28, 2012 at 8:42 AM, Brendan Churchill <Brendan.Churchill@utas.edu.au> wrote:
>
>> I am using some ordinal variables, which have some numeric missing values, in a multilevel model. In some previous research, I have seen researchers include a 'Missing' independent variable in their model to account for some of the 'missingness' - or rather to control for the missing values, but I don't quite understand how to do it in Stata or even if that's a good way to do it. I've tried to make a binary variable in which the missing values are coded 1 and the rest of the values are coded 0 but the model rejects this because it's collinear.
>>
>> Is this how you do it? Or is there a variable for the entire data set that is created to account for all missing variables?
>>
>> I'd great appreciate any advice or assistance anyone out there could 
>> provide

*
*   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/

================= Disclaimer: The content of this email message and any attachments are intended for specific recipients and are privileged and confidential. If you have received this email in error, please kindly delete its contents and advise the Refugee Processing Center (RPC) via reply e-mail of the delivery mistake.

*
*   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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index