Rita Luk
> I observed after I tsset PANEL TIME in my data, I can xtdes
> without setting
> iis tis. Is this true for all xt commands? Is there any difference in
> using tsset instead of iis tis in xt commands ?
On 19 April Vince Wiggins of Stata Corp posted this:
Stephen Jenkins <[email protected]> asks how -tsset- relates to -iis- and
-tis-,
> My experiments suggest that
> -tsset panelvar timevar-
> seems very similar to
> iis panelvar
> tis timevar
>
> [Try -tsset panelvar timevar-, as above, then -iis- and -tis-.]
> I discovered this having -tsset- my data and then run -xtlogit-
> successfully, without having first done a -iis- or -tis- or used the
> equivalent options in -xtlogit- .]
>
> To what extent are they really similar?
> E.g. I note from the Manual that -tsset- leaves behind r(timevar) and
> r(panvar) whereas -iis- and -tis- set characteristics _dta[iis] and
> _dta[tis]. Any reason for this difference that is useful to know about?
Stephen is right, there is quite a bit of overlap between -tsset- and the
pair
of commands -iis- and -tis-.
-tsset- was introduced in Stata 6 along with time-series operators, and it
provides several new features in addition to subsuming most of the the role
of
the -iis/tis- commands. Those new features include enabling time-series
operators, checking the data to be sure that it does not contain duplicate
time periods, and setting the periodicity of the data, e.g. quarter, yearly,
etc. In general, -tsset- is always preferred to -tis- and often to -iis-,
see
below.
The -iis- and -tis- command have been in Stata much longer. -iis- has
always
been the workhorse because all of the -xt- commands must know the variable
that stores the panel/group identifier. Its sibling, -tis-, was introduced
to
provide a similar mechanism for the few -xt- commands that also needed a
variable identifying time, these commands include -xtgls- and -xtgee-.
Stephen notes that -iis- and -tis- set the characteristics _dta[iis] and
_dta[tis]. In truth, that is about all they do, it is the -xt- commands
that
then use this information. -tsset- also sets these characteristics and sets
some additional characteristics used internally by the time-series
operators.
The commands also clear each others settings. So, typing -iis id- will
clear
any existing -tsset- settings, and the reverse is also true.
All of the -xt- commands that recognize the -iis- and -tis- settings, or
that
take the i() and t() options, respect the settings of -tsset-. When you
have
data with both group/panel and time identifiers, -tsset- is preferred
because
it enables time-series operators. The relationship between -tsset- and
-iis/tis- should be clearer in the documentation.
I still use -iis-, or the i() option, because many of the -xt- commands
require only a panel/group variable; they do not need a time variable, for
example -xtlogit-. (-tsset- requires a time variable and optionally allows
a
panel/group variable). On the other hand, I almost never use tis except in
contracted form, " `tis the rare day that I use -tis-."
Nick
[email protected]
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
