re: st: tsset error codes

[Christopher Baum <kit.baum@bc.edu>]

<>
I said

If I'm not mistaken, CES data are a cross-section for each quarter, so you do not have a time series; you have as many 'repeated time values' as there are individuals in a quarterly dataset. Thus you cannot use tsset with one argument no matter what. But it's not a panel either, in that households rotate out of the survey after 5 quarters. So you have to treat these data with caution.


'With caution' means if each household has a unique identifier (something akin to the social security # of the head of household, which of course it will not be), but households are in the sample for 5 quarters and then rotate out, you could do

tsset hhid qtr
or
xtset hhid qtr

What you do NOT want to do is compute some measure across time (such as a lag or difference) that will pick up values from different households. So you do NOT want to do something like number the households sequentially in each quarter and use that as hhid. That would not recognize that there are 5 obs. for each household over time, nor would it make clear that hhid=126 at one point in time could be a different unit than hhid=126 in the next period.

Kit


Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                             An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
  An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html


*
*   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/