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]

From |
Christopher Baum <kit.baum@bc.edu> |

To |
"statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |

Subject |
Re: Re: st: Constructing a variable from standard deviations |

Date |
Tue, 23 Nov 2010 08:29:56 -0500 |

<> On Nov 23, 2010, at 2:33 AM, MP wrote: > The approach used by Kit seems to be familiar to an approach i used > earlier. My prof gave me the advice to include all the dummies in the > regression and then calculate the residual per occupation, since this is > what Bonin et all do. > > However, you guys claim that the estimates from this procedure would be > meaningless. I will start over again and make regression per occupation > and then predict the residuals. I wouldn't say meaningless. It is just a classic case of groupwise heteroskedasticity. Running that single regression with dummies, you could then use -robvar- to examine the hypothesis that the residual variance is constant, or differs, across observations. See "Stata Tip 38: Testing for groupwise heteroskedasticity", Stata Journal, 2006, 6:4, 590-592, freely available from IDEAS (http://ideas.repec.org) or the SJ website. The point others have made is that if you believe from the outset that the errors in this model are groupwise heteroskedastic, then why not allow for that in the first place? A procedure such as I suggested, in which you estimate the model separately for each group, compute the residuals, and then compute their sd is more in line with that assumption. It also allows the coefficients on other variables to differ across groups, which is the primary difference between this approach and the single-regression-with-dummies approach. In the latter, you are assuming that the effects of the regressors are constant across groups. That is a testable hypothesis, and I would not be surprised if you could easily reject it. 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/

- Prev by Date:
**Re: st: Problems with ML: misclassification-adjusted oprobit** - Next by Date:
**st: xi and i. command** - Previous by thread:
**re: Re: st: Constructing a variable from standard deviations** - Next by thread:
**st: Removing Spaces from Variable Labels** - Index(es):