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

From |
"Martin Weiss" <martin.weiss1@gmx.de> |

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

Subject |
Re: st: Imbalance in control versus treated group, and weights |

Date |
Thu, 9 Oct 2008 22:59:18 +0200 |

http://www.stata-journal.com/article.html?article=st0136 HTH Martin _______________________

To: <statalist@hsphsun2.harvard.edu> Sent: Thursday, October 09, 2008 10:56 PM Subject: SV: st: Imbalance in control versus treated group, and weights

I have another question. I followed the advice and looked into propensityscore reweighting (PSR) and regression discontinuity (RD). Google pointedme to Austins presentation about this topis,http://www.stata.com/meeting/6nasug/causal.pdfI have read through the presentation, but I do not understand all theassumptions that underpins RD. My problem pass the first assumption thatmy treatment is not randomly assigned, though it started out as arandomized controlled trial, just that not all those supposed to have atreatment got one. Further, the assignment variable is based on aobservable variable. Or well, it was not supposed to be an assignmentvariable, but it turned out to be, and consequently contaminated thetreated versus the control group.However I am uncertain what the second assignment is telling me, quotingAustins presentation"The crucial second assumption is that there is a discontinuity at somecutoff value of the assignment variable in the level of treatment."My assignment variable do produce a jump in the level of treatment, but Iam unsure whether this actually means that I pass assumption 2?I also downloaded the RD package from SSC (findit regressiondiscontinuity). However, I am still unclear how I can relate the providedexample to my own problem. I am having trouble locating other examples,and any tip would be greatly appreciated.Best wishes, Alexander Severinsen -----Opprinnelig melding-----Fra: owner-statalist@hsphsun2.harvard.edu[mailto:owner-statalist@hsphsun2.harvard.edu] På vegne avAlexander.Severinsen@telenor.comSendt: 8. oktober 2008 19:11 Til: statalist@hsphsun2.harvard.edu Emne: SV: st: Imbalance in control versus treated group, and weights Thank you for the advice. Very helpful!In this spesific case z is a dummy, and if z=1 then this will increase thelikelihood of observing x=1. And yes, I do observe outcomes for the groupthat was supposed to be treated, but were not.Best wishes, Alexander -----Opprinnelig melding-----Fra: owner-statalist@hsphsun2.harvard.edu[mailto:owner-statalist@hsphsun2.harvard.edu] På vegne av Austin NicholsSendt: 8. oktober 2008 18:39 Til: statalist@hsphsun2.harvard.edu Emne: Re: st: Imbalance in control versus treated group, and weightsIt is possible that some kind of propensity score reweighting orregression discontinuity design would be appropriate here, but withoutmuch more information, it is hard to offer any specific advice. How doesz affect x in the group supposed to have x=1? Do you observe outcomes forthe group supposed to have x=1 but having x=0? Etc.Running a probit with the assumption E(y)=F(b0+b1*x+b2*z) seems unlikelyto recover a good estimate of the effect of x on y unless that assumptionis actually true!On Wed, Oct 8, 2008 at 12:23 PM, <Alexander.Severinsen@telenor.com>wrote:Dear Statalisters,I have the following problem. I have given a sample of 10000 people astargets for receiving an offer, and I have a control group equal to 5000people. I know that the potentially treated and the controlgroup isrepresentative. However, without my knowledge only 8000 of the 10000targets were treated, and a specific criteria was used to pick those 8000from the 10000.This has created an imbalance between my controlgroup and those treated,and this imbalance is identified and only concerns one variable. I wantto investigate whether the offer given could reduce the defection rate ofcustomers, but the variable that created this imbalance is known tohugely impact the defection rate. To reduce this problem I would like touse weights in Stata, but I am unsure on how to approach this? Any tipswould be greatly appreciated.Also, say that I did not correct for this, and did the following probitmodel with the following variables, y=defected/not defected,x=treated/control, z=factor that created imbalance:y=b0+b1*x+b2*zwould it be appropriate to say that it was possible to control for theimbalance by including it as a independent variable in this fashion?Best wishes, Alexander Severinsen* * 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/ * * 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/ * * 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/

* * 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/

**Follow-Ups**:**SV: st: Imbalance in control versus treated group, and weights***From:*<Alexander.Severinsen@telenor.com>

**References**:**st: Imbalance in control versus treated group, and weights***From:*<Alexander.Severinsen@telenor.com>

**Re: st: Imbalance in control versus treated group, and weights***From:*"Austin Nichols" <austinnichols@gmail.com>

**SV: st: Imbalance in control versus treated group, and weights***From:*<Alexander.Severinsen@telenor.com>

**SV: st: Imbalance in control versus treated group, and weights***From:*<Alexander.Severinsen@telenor.com>

- Prev by Date:
**SV: st: Imbalance in control versus treated group, and weights** - Next by Date:
**Re: st: estimating distribution functions from Kernel densities** - Previous by thread:
**SV: st: Imbalance in control versus treated group, and weights** - Next by thread:
**SV: st: Imbalance in control versus treated group, and weights** - Index(es):

© Copyright 1996–2017 StataCorp LLC | Terms of use | Privacy | Contact us | What's new | Site index |