Bookmark and Share

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]

Re: Re: st: psmatch2: ATT as percentage change using logs?


From   Lukas Borkowski <[email protected]>
To   [email protected]
Subject   Re: Re: st: psmatch2: ATT as percentage change using logs?
Date   Thu, 15 Aug 2013 17:10:02 +0200

I am a bit confused here. Ubydul, did you suggest to use a poisson regression (thanks Nick for spelling it out) to estimate the ATT in terms of a percentage change in the outcome variable?

On 15.08.2013, at 16:06, Nick Cox <[email protected]> wrote:

> The poison model sounds exactly right here, but do check your
> presentation slides, paper drafts, etc. for this typo before going
> public, otherwise even your friends will never let you hear the end of
> this.
> 
> You used a Poisson model.
> 
> (Also, more fundamentally: Please don't start new threads by replying
> to old ones _and without even changing the title\=, as is explained in
> the FAQ.)
> 
> Nick
> [email protected]
> 
> 
> On 15 August 2013 14:58, Ubydul Haque <[email protected]> wrote:
> 
>> I used poison regression model to explore the association of indoor
>> residual spraying (IRS) and long lasting insecticide treated net
>> (LLIN) and malaria cases.
#
Lukas Borkowski
University of London, School of Oriental and African Studies (SOAS)

M: [email protected]



On 15.08.2013, at 15:58, Ubydul Haque <[email protected]> wrote:

>> Hello,
>> I used poison regression model to explore the association of indoor
>> residual spraying (IRS) and long lasting insecticide treated net
>> (LLIN) and malaria cases. The outcome is the following:
>> xi: xtpoisson   totalcases  pc_llin prop_irs i.year,  irr exp(
>> population) i( district_code) i.year
>> 
>> 
>> ------------------------------------------------------------------------------
>> malaria cases |        IRR   Std. Err.      z    P>|z|     [95% Conf. Interval]
>> -------------+----------------------------------------------------------------
>>     per capita_llin |   1.185207   .0483431     4.17   0.000
>> 1.094145    1.283848
>>    proportion_irs |   .7504642   .0696331    -3.09   0.002
>> .6256765 .9001402
>> _Iyear_2009        |   .7955212   .0128567   -14.15   0.000
>> .7707175 .8211232
>> _Iyear_2010       |   .6376188   .0165131   -17.38   0.000
>> .6060614 .6708194
>> _Iyear_2011       |   .0790123   .0023992   -83.59   0.000
>> .0744472 .0838574
>> _Iyear_2012       |    .013673   .0008019   -73.19   0.000
>> .0121883 .0153386
>>  population        | (exposure)
>> -------------+----------------------------------------------------------------
>>    /lnalpha |   1.204061   .2083535                      .7956957  1.612426
>> -------------+----------------------------------------------------------------
>>       alpha |   3.333627   .6945728                      2.215982   5.014964
>> ------------------------------------------------------------------------------
>> Likelihood-ratio test of alpha=0: chibar2(01) =  2.6e+04 Prob>=chibar2 = 0.000
>> 
>> http://www.ats.ucla.edu/stat/stata/output/stata_poisson_output.htm
>> 
>> 
>> The probable explanation is "An additional increase IRS in coverage by
>> one percent will reduce malaria by a factor of XX (recalculate the
>> 0.75 by using percentages rather than proportion), while holding LLINs
>> in the model constant. On the other hand every additional ITN per
>> capita will increase malaria by a factor of 1.19, while holding IRS in
>> the model constant".
>> 
>> Readers are interested to know the real impact of IRS and ITN to
>> reduce malaria cases. But can I say like this "IRS coverage was
>> associated with ....%/proportion/factor of reduced malaria prevalence
>> while ITN was associated with %/proportion/factor of increased malaria
>> prevalence".
>> 
>> Thank you.
>> 
>> Ubydul Haque
>> Post Doctoral Fellow
>> Johns Hopkins University
>> 615, North Wolfe Street
>> Baltimore, Maryland
>> 21218, USA
>> 
>> On Thu, Aug 15, 2013 at 9:25 AM, Lukas Borkowski <[email protected]> wrote:
>>> Dear statalist,
>>> 
>>> I am using the user-written command -psmatch2- (written by Leuven and Sianesi, 2003) to estimate the ATT of my participation variable on several outcome indicators. I am also using a DiD-regression approach to estimate the ATT where I transformed some of the indicators into logs to see the percentage change caused by the treatment. I wonder whether it is also reasonable to transform the outcome indicators into logs when running -psmatch2-. Can someone give me a quick hint? Thanks!
>>> 
>>> Best, Lukas
>>> 
>>> Leuven, E., and Sianesi, B. (2003) PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing. Available at: http://ideas.repec.org/c/boc/bocode/s432001.html
>>> 
>>> #
>>> Lukas Borkowski
>>> University of London, School of Oriental and African Studies (SOAS)
>>> 
>>> M: [email protected]
>>> 
>>> 
>>> 
>>> 
>>> *
>>> *   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/
>> 
>> 
>> 
>> -- 
>> Ubydul Haque
>> Post Doctoral Fellow
>> Johns Hopkins University
>> 615, North Wolfe Street
>> Baltimore, Maryland
>> 21218, USA
>> 
>> *
>> *   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/
> 

*
*   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index