Statalist


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

Re: st: weights panel-survey data


From   mdeidda@stern.nyu.edu
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: weights panel-survey data
Date   Tue, 14 Oct 2008 17:02:32 -0400

thanks a lot!
Manuela

Steven Samuels wrote:

Manuela, If the public-use data does not identify the original clusters, the survey distributors almost certainly included other variables which would allow you to compute standard errors. Check the survey documentation. You will probably see mention of: 1) (pseudo)stratum and (pseudo) cluster variables; or 2) "replicates" and "replicate weights", for example "bootstrap" or "jackknife" replicates.

Importance weights?  Do not use them! See -help- for "weights".

-Steve


On Oct 14, 2008, at 4:00 PM, mdeidda@stern.nyu.edu wrote:

Dear Steve, thanks a lot...actually data are clustered by province, but this information is not available to the public. By the way I have discovered that many of the commands I am using in Stata only allow importance weights. I have read that using importance weights is not useful to provide a correct estimation...do you agree with that? I am a little bit confused, I tried to do the same regression using probability weights and importance weights, and the coefficient estimates are the same, but the number of observations and the standard error is different.
Do you think I shoud use importance weights?
thanks a lot,
Manuela

Steven Samuels wrote:


I had another thought. Your survey design may have included multi- stage sampling and stratification. -xtreg- cannot accommodate clusters other than panels. Within Stata you have one choice for an analysis that accommodates weights and clusters: - gllamm-.

-Steve

On Oct 14, 2008, at 2:57 PM, Steven Samuels wrote:

Manuela, You state you have a 3 year panel. Does this mean that each household was followed for three years or that you are examining three calendar years?

I suggest that you give each household the weight it had for the first time it was selected. If all households were selected in the first calendar year, then the weights represent the population of that year. If households

Manuela, You state you have a three-year panel. Does this mean that each household was followed for three years or that you are examining three calendar years?

I suggest that you give each household the weight it had for the first time it was selected. If all households were selected in the first calendar year, then the weights represent the population of that year. If households rotated in and out, they represent the year in which they came in. The total sample then represents the "household-years" of the survey period. This is Austin Nichols's formulation. (http://www.stata.com/ statalist/archive/2007-11/msg00245.html).

There is a complication if you are interested in three calendar years, say 2000 2001 2002, but some households started follow-up before 2000 and rotated out after 2000. Give those HH the weight for 2000.


-Steve


On Oct 14, 2008, at 1:45 PM, mdeidda@stern.nyu.edu wrote:
Dear Steven, the panel is unbalanced, the sampling unit is the household. The survey is carried on biannually. To each household is attributed a probability weight each year...but weights are modified each year in order to take into consideration changes in some known characteristics of the population.
thanks a lot

Steven Samuels wrote:

What are your panels? Do they all have three years of data? Please describe the population sampling process. How did the sampling produce different weights between years?

-Steve


On Oct 14, 2008, at 12:21 PM, mdeidda@stern.nyu.edu wrote:

Dear all
I am estimating a 3 year panel (random effect) using survey data. To get correct estimates I should use sampling weights, but the command xtreg, re does not allow me to use weights...I can use xtreg with the option mle, which allow me to use analytic weights. But stata requires weights to be constant within panels. Is it correct to use the weights of the last year, and assume they are constant within panel?
Or should I simply ignore weights?
I really appreciate your help,
Manuela


*
*   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/


*
*   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/


*
*   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/
rotated in and out, they represent the year in which they came in. The total sample then represents the "household-years" of the survey period. This was Austin Nichols's formulation. (http://www.stata.com/ statalist/archive/2007-11/msg00245.html).

There is a complication if you are interested in three calendar years, say 2000 2001 2002, but some households started follow-up before 2000 and rotated out after 2000. Give those HH the weight for 2000.

If you are analyzing a national survey data set, the survey designers may have created weights for combining data from different years. So I also suggest that you check the survey documentation.

-Steve


On Oct 14, 2008, at 1:45 PM, mdeidda@stern.nyu.edu wrote:

Dear Steven, the panel is unbalanced, the sampling unit is the household. The survey is carried on biannually. To each household is attributed a probability weight each year...but weights are modified each year in order to take into consideration changes in some known characteristics of the population.
thanks a lot

Steven Samuels wrote:

What are your panels? Do they all have three years of data? Please describe the population sampling process. How did the sampling produce different weights between years?

-Steve


On Oct 14, 2008, at 12:21 PM, mdeidda@stern.nyu.edu wrote:

Dear all
I am estimating a 3 year panel (random effect) using survey data. To get correct estimates I should use sampling weights, but the command xtreg, re does not allow me to use weights...I can use xtreg with the option mle, which allow me to use analytic weights. But stata requires weights to be constant within panels. Is it correct to use the weights of the last year, and assume they are constant within panel?
Or should I simply ignore weights?
I really appreciate your help,
Manuela


*
*   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/




*
*   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/




*
*   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/



*
*   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/



*
*   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/



*
*   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/


*
*   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/


*
*   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   |   What's new   |   Site index