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, is already up and running.

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

Re: st: lincom command

From   Ari Samaranayaka <>
To   <>
Subject   Re: st: lincom command
Date   Wed, 13 Mar 2013 10:54:57 +1300

Thank you Maarten for the comment. Your idea was my very first try, it was given up later for the difficulty of interpreting results. What you are saying is to treat A,B,C,D as 4 separate binary variables. Problem with that is overlap between those binary variables is high, therefore colinearity. For this reason estimated IRR for ethnicity A and B can be quite different eventhough overlap of people between ethnicities A and B are quite large. In that situation I am unable to interpret (understand) estimates.

On 12/03/2013 10:23 PM, Maarten Buis wrote:
The much easier solution is to just create separate indicator
variables for whether or not a person feels (s)he belongs to ethnicity
A. The full set of indicator variables will not be mutually exclusive,
but that is exactly what you want, and you will get directly the IRRs
you are looking for without using any post-estimation commands.

-- Maarten

On Tue, Mar 12, 2013 at 5:36 AM, Ari Samaranayaka
<> wrote:
Hi folks
I need to use a poisson multivatiate model using a human health dataset,
outcome is a specific health outcome, one of the explanatory variables is
ethnicity. One of the results I need to estimate is IRR for various
ethnicities relative to a reference ethnicity. Let us say ethnicities are A,
B, C, D. Some people belong to multiple ethnicities, therefore ethnicity is
not a variable with mutually exclusive categories. For this reason I cannot
represent it using dummy variables. Does any one know how to represent a
categorical variable when categories are not mutually exclusive?

What I have done is, have created a set of mutually exclusive ethnic
categories so that i can use them in the model. Say those categories are P,
Q, R, S, T (here I have more groups than above, I have no research interest
on them). People from single ethnicity in original ethnicity classification
now belongs to multiple categories in new classification, but new
classification can be represented by dummy variables. Then I can have
estimates (regression coefficients and IRR) for each of these new ethnic
categories, but what I really need is the estimates for my original ethnic
categories. Does any one know how to convert estimates for P, Q, R, S, T
into estimates for A, B, C, D?

I thought I can use stata lincom command for that as stata documentation
says. For example, required linear combination for ethnicity A is determined
by the distribution of ethnic A people across ethnic groups P to T. I know
those distributions for all interested ethnicities. Do you think this is a
correct  approach?

Thank you in advance for any help.

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index