Kit,
Thanks very much for the explanation. That was helpful. You were
certainly correct on the coffee comment. I had already cribbed the
`eqn' logic from your Modern Econometrics text and run with manually
generated lags. Using the tsset will be much cleaner. Thanks again.
Tom
On Mon, Oct 13, 2008 at 4:15 PM, Kit Baum <[email protected]> wrote:
> < >
> Thomas said:
>
> I am trying to use sureg on a dataset of approximately 50 firms with
> 900 trading days of observations:
>
> . xtset
> panel variable: CompanyNum (strongly balanced)
> time variable: TradeDateNum, 501 to 1402
> delta: 1 unit
>
> I keep only the necessary variables:
>
> . keep TradeDateNum CompanyNum LnRt*
>
> and reshape from long to wide for the company specific variables:
>
> . reshape wide LnRtFiveYrMid LnRtStock, i(TradeDateNum) j(CompanyNum)
> (note: j = 1 2 3 4 5 6 8 9 10 11 12 13 14 15 16 17 18 19 20 22 23 24
> 25 26 27 28 29 30 32 33
>> 34 35 36 37 38 39 40 41 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57)
>
> Data long -> wide
> -----------------------------------------------------------------------------
> Number of obs. 47806 -> 902
> Number of variables 10 -> 113
> j variable (53 values) CompanyNum -> (dropped)
> xij variables:
> LnRtFiveYrMid -> LnRtFiveYrMid1
> LnRtFiveYrMid2 ... LnRtFiveYrM
>> id57
> LnRtStock -> LnRtStock1 LnRtStock2
> ... LnRtStock57
> -----------------------------------------------------------------------------
>
> I then attempt to execute sureg on my current pooled OLS specification
> for just two firms:
>
> . sureg (LnRtFiveYrMid1 L(1/2).LnRtFiveYrMid1 L(0/4).LnRtCDX LnRtCMT
> LnRtTED L(0/2).LnRtStoc
>> k1) (LnRtFiveYrMid2 L(1/2).LnRtFiveYrMid2 L(0/4).LnRtCDX LnRtCMT LnRtTED
>> L(0/2).LnRtStock2
>> )
> variable CompanyNum not found
>
> and I get the error CompanyNum (the dropped variable after reshaping
> to wide) not found. So far, the only way it will run is if I exclude
> time series operators:
>
> . sureg (LnRtFiveYrMid1 LnRtCDX LnRtCMT LnRtTED LnRtStock1)
> (LnRtFiveYrMid2 LnRtCDX LnRtCMT
>> LnRtTED LnRtStock2)
>
> Seemingly unrelated regression
> ----------------------------------------------------------------------
> Equation Obs Parms RMSE "R-sq" chi2 P
> ----------------------------------------------------------------------
> LnRtFiveY~d1 887 4 .0453147 0.1464 152.27 0.0000
> LnRtFiveY~d2 887 4 .0479444 0.0777 74.68 0.0000
> ----------------------------------------------------------------------
>
> ------------------------------------------------------------------------------
> | Coef. Std. Err. z P>|z| [95% Conf.
> Interval]
> -------------+----------------------------------------------------------------
> LnRtFiveY~d1 |
> LnRtCDX | .5369359 .054891 9.78 0.000 .4293515
> .6445204
> LnRtCMT | -.3181771 .0979303 -3.25 0.001 -.5101169
> -.1262374
> LnRtTED | -.0106727 .0188642 -0.57 0.572 -.0476459
> .0263005
> LnRtStock1 | .0629164 .1021791 0.62 0.538 -.1373509
> .2631838
> _cons | -.0006975 .0015239 -0.46 0.647 -.0036843
> .0022893
> -------------+----------------------------------------------------------------
> LnRtFiveY~d2 |
> LnRtCDX | .362548 .0568002 6.38 0.000 .2512216
> .4738745
> LnRtCMT | -.3034092 .1032472 -2.94 0.003 -.5057699
> -.1010485
> LnRtTED | -.0484131 .019944 -2.43 0.015 -.0875027
> -.0093235
> LnRtStock2 | -.0617068 .0777801 -0.79 0.428 -.214153
> .0907394
> _cons | -.0000496 .0016117 -0.03 0.975 -.0032085
> .0031093
> ------------------------------------------------------------------------------
>
> Does sureg prevent the use of time series operators? If so, am I best
> to manually generate all the lags before reshaping or after? Any
> other suggestions? Thanks.
>
>
>
> The problem is that -tsset- (or -xtset-) remembers how your original
> long-format data were defined, in terms of CompanyNum and TradeDateNum.
> -sureg- is happy to accept timeseries operators, as -help sureg- indicates.
> But you must after using -reshape- reestablish the -tsset-, which will now
> be just
>
> tsset TradeDateNum
>
> as what you have now is a pure timeseries dataset. Then you should be able
> to give the -sureg- command.
>
> As you have 57 equations to specify, I suggest you use a -local- and a
> -forvalues- loop over company numbers to build up the equation list.
> For instance, as a simplified example,
>
> local eqlist
> forv i = 1/57 {
> local eqlist "`eqlist' (LnRtFiveYrMid`i' LnRtCDX LnRtCMT LnRtTED
> LnRtStock`i')"
> }
>
> Just make sure you never use an equals sign in this local statement. You can
> then say
>
> sureg `eqlist'
>
> and go have a cup of coffee.
>
>
> Kit Baum, Boston College Economics and DIW Berlin
> http://ideas.repec.org/e/pba1.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/
>
--
Thomas Jacobs
*
* 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/
