Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Carlo Lazzaro" <carlo.lazzaro@tin.it> |
To | <statalist@hsphsun2.harvard.edu> |
Subject | st: R: Logarithmic transformation in psm results |
Date | Thu, 17 Mar 2011 12:14:01 +0100 |
Sergio wrote: "... using the original variable expressed in monetary units the value corresponding to treated and controls not coincide with the transformed values of the results in logarithms". Without other details, and provided that your problem concerns transformation from the log scale only, the first advice is to check that, when back transforming to the original metric (I would assume that it is arithmetic),the sample mean is actually obtained via the following formula: arithmetic mean=exp(ln_mean+ln_variance/2) where: ln_mean is the mean on the log scale; ln_variance is the variance on the log scale. Kindest Regards, Carlo -----Messaggio originale----- Da: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Per conto di s Inviato: giovedì 17 marzo 2011 11.55 A: statalist@hsphsun2.harvard.edu Oggetto: st: Logarithmic transformation in psm results Dear statalisters, I'm estimating propensity score matching (with psmatch2 routine) for a sample of firms participating in a program. My outcome variable (turnover of the firm) is expressed in logarithm, and the results are similar to these: ---------------------------------------------------------------------------- ------------------------------------------ Sample Treated Controls Difference S.E. T-stat ----------------------------+----------------------------------------------- ----------------------------------------- Unmatched | 13.2005389 8.90516337 4.29537557 .323484337 13.28 ATT | 12.9789829 12.284779 .694203895 .303516206 2.29 ----------------------------+----------------------------------------------- ----------------------------------------- How can I transform the results in order to interpret this results in terms of the original values. My doubt arise mainly because when I estimate the ATT using the original variable expressed in monetary units the value corresponding to treated and controls not coincide with the transformed values of the results in logarithms. ---------------------------------------------------------------------------- -------------------------------------------------- Sample Treated Controls Difference S.E. T-stat ----------------------------+----------------------------------------------- ------------------------------------------------ Unmatched | 3940905.5 642598.305 3298307.2 266059.508 12.40 ATT | 2687231.69 2091249.89 595981.797 631105.497 0.94 ----------------------------+----------------------------------------------- ------------------------------------------------ Thanks in advance for your help, Sergio * * 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/