Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: running simple regression on mixed data of cross-section and panel


From   Steven Samuels <[email protected]>
To   [email protected]
Subject   Re: st: running simple regression on mixed data of cross-section and panel
Date   Mon, 6 Dec 2010 13:29:38 -0500

Without going into the details of your particular data set, I urge you to consult the study documentation . Weighting for such rotating studies really requires a very clear definition about what population each individual represents. You might well have to construct new weights to do your analysis. Perhaps the study documentation ,or publications based on the survey, will provide guidance; Otherwise, you will need to write to the survey organization for advice.

Steve

Steven J. Samuels
[email protected]
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783

On Dec 6, 2010, at 1:12 PM, Melaku Fekadu wrote:

Hi Steve,

Thanks for your help and here is a detail about the survey structure.

The labor force survey is quarterly. In a given quarter, no individual
appears more than once in the data. It is representative of the entire
population and employment figures can be drawn from it.

The survey returns to the same individuals in the subsequent quarters
to be able to follow up their employment situation.  The panel has a
rotating structure: not all panels start at the same time. An example
is given below. All in all, an individual should be interviewed in 4
quarters: in the first two consecutive quarters and in two more
quarters after a spell of two quarters (with no interviews conducted)
in between.

For example, focusing about individuals who are interviewed in Q1
2009, as seen under Q1 bellow, they are composed of four types of
individuals:

1) those who completed their final (fourth) interview in Q1. These
individuals did their last 3 interviews in previous years
2) those who do their third interview in Q1. These individuals will do
their final (fourth) interview in the next quarter, Q2, this year.
They did their last 2 interviews in previous years.
3) those who do their second interview in Q1. These individuals will
have a spell of 2 quarters (Q2 and Q3) with no interviews and will do
their third interview in Q4, this year. They will do their last
interview in Q1 of next year, 2010.
4) those who are interviewed for the first time in Q1. These
individuals will do their second interview in Q2, this year. They will
take a spell of two quarters (Q3 and Q4) with no interviews conducted.
And then they will do their third and fourth interviews in the Q1 and
Q2 of next year, 2010.

Individuals who are interviewed in the other quarters of 2009 are also
composed of four types of individuals with similar features.

The problem of recurring individuals arises when I use a pooled data
from 4 quarters. As we see the recurrence of respondents is not
accidental. As seen bellow only 2/16 or 12.5% of the observations in
the pooled data appear only once (those who are interviewed in Q1
their last – fourth – interview, and those who are interviewed in Q4
their first interview). The rest 87.5% appear twice.

I would like to do my regressions in the pooled data of four quarters
and I am afraid that this recurring feature of individuals has an
impact on my regressions. My question is how do I take care of it?
Does 'xtreg' helps? If so, I would appreciate any help on how to use
xtreg.

Year=2009
Q1      Q2        Q3      Q4

4
3        4
        3        4
                 3        4
2                          3
1       2
       1         2
                 1        2
                          1




On Sun, Dec 5, 2010 at 12:38 AM, Steven Samuels <[email protected]> wrote:
--

A panel survey is one which deliberately resurveys the same units. My guess is that you are talking about repeated independent samples of a population at different times. If not, please provide more details. If so, the chance appearance of individuals in more than one survey is not an issue. Use Stata survey commands, and incorporate "survey occasion" into the stratum definition for the -svyset- command. For most repeated surveys, the study
documentation will give advice about this kind of issue.

Steve

Steven J. Samuels
[email protected]
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783

On Dec 3, 2010, at 2:35 PM, Melaku Fekadu wrote:

Dear all,


I want to do a regression of employment (logistic) and wage
(loglinear) on some demographic variables. I do the regressions on a
cross-section data of labor force survey. However, small proportion of
individuals in the data recur more than once.


How do I take into account the fact some people appear more than once
in the data. Do I have to be concerned from running a simple
regression?


Thanks
fekadu,
*
*   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index