Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


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

st: Regression output: Nwest and Probit


From   Daniel Park <danielpark87@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: Regression output: Nwest and Probit
Date   Mon, 17 Jun 2013 11:42:03 +0200

Hello everyone,

before I start I ll apologize for any noob-ish questions. I previously
worked with Eviews for my master thesis until I had to realize that
the things I need to do can't be done with Eviews.

So since yesterday I have been trying to teach myself some STATA.
For my thesis, I am performing forecasts using a probit model. Due to
the overlapping nature of the forecasts, I would like to correct the
serial correlation of the errors by using Newey-West techinique.
Delta is one quarter and I want to do predictions 1 to 8 quarters ahead.

What I got so far:

tsset date
time variable:  date, 1900q1 to 2013q1
delta:  1 quarter

nwest probit dependentvar independentvar, lag(X) t(date)

first annoying thing is that the lag operator "L" doesn't work with
nwest, or at least i am too stupid to do so. Is there any way except
creating new time series with manually lagged independent variables?
If i use "probit" only, the lag operator works perfectly fine.

Also, i set X equal the lag of the independent variable minus 1 (so if
the independent variable has a lag of two quarters I set X equal 1). I
thought that for one quarter ahead I don't yet need to correct for
serial correlation,or am I mistaken?

My question is concerning the output that nwest delivers. Eviews
outputs a nice table with all sorts of information for any sort of
regression (problem with Eviews was that there is no function to
calculate robust SE based on Newey West for probit models). However,
the nwest command only outputs

Regression with Newey-West standard errors
maximum lag : 0
Number of obs  = 251
F(  1,   249) = 13.13
Prob > F = 0.0004

------------------------------------------------------------------------------
                |                      Robust
     ptinrec |      Coef.         Std. Err.       t      P>|t|
[95% Conf. Interval]
-------------+----------------------------------------------------------------
  dlog_djind |  -1.888772   .5211739     -3.62   0.000      -2.915243
  -.8623005
       _cons |    .218619     .0268878     8.13    0.000     .1656625
   .2715755
------------------------------------------------------------------------------

If it comes out distorted, its basically just the coefficients with
the robust SE and the t-statistics together with probability and 95%
confidence interval.

Is there any option to extend this table automatically to include
basic regression output like R-squared, adjust R-squared, the SE of
the regression, values of the log likelihood etc?

Thanks in advance
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index