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 |
Steve Samuels <sjsamuels@gmail.com> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: Power estimation in Cox ph model |

Date |
Wed, 29 Feb 2012 17:13:45 -0500 |

Sergey, See Version 12 manual shows for example 5, page 286. the number before the comma is the coefficient you wish to detect. It is the log of the hypothetical HR you wish to detect. In stpower cox 1, ... 1 was the log(hr) to detect. HR = exp(1)= 2.718 or exp(-1) 1/2.718. It has no relevance to your problem; it was just an illustration of the technique. Take values for the HR from previous studies. You quoted an HR = 1.20. So I'll presume that is one the values from a previous study. Then the beta coefficient to detect is ln(1.20) = 0.1823. Also your failure probability = 53/2387 = 0.0222. n() contains the number of subjects. You don't say what the SD of log hydrazine was, but suppose it was, e.g. 2.1 ************************************************ stpower cox 0.1823, n(2387) sd(2.1) failprob(0.0222) ************************************************ For the adjusted analysis, you will have to add the r-squared for the correlation of each chemical with internal radiation and all other covariates. You can guess that or, if you have the data, do your own multiple regression of, e.g. log(hydrazine exposure) on the other factors and use that r-square. If, for hydrazine t r-square = .04, the statment would be:. ************************************************ stpower cox 0.1823, n(2387) sd(2.1) r2(.04) failprob(0.0222) ************************************************ So you can check more than one set of assumptions at a time and produce a table. Suppose that two another plausible values for hydrazine are hr = hr = 1.25, with coefficient beta ln(1.25) = .22314355. and hr = 1.5 with beta = .40546511 Then you can get a table by running: ************************************************ stpower cox (0.1823 0.2231 0.4055), n(2387) sd(2.1) r2(.04) failprob(0.0222) ************************************************ Best of luck, Steve sjsamuels@gmail.com On Feb 29, 2012, at 9:17 AM, ZHIVIN Sergey wrote: Dear Steve and all Statalist, Good to know about Stata name, I always wrote in capitals taking in mind the examples of SAS and SPSS...Sorry! I have read a manual on -stpower-, but I still have questions. Could you help me please? As I wrote our study is historical cohort, and now I want to estimate power of Cox model between exposure to hydrazine and lung cancer risk. So, we have 53 deaths in 2387 workers (however, not all workers are exposed). I obtained sd of log for cumulative exposure to hydrazine, and failprob=53 deaths/2387 workers. HR is set to 1.20... Does my formula look correct? stpower cox, failprob(), sd() n(2387) hratio (1.20) Or should I take not all workers as n? I also found the models with stpower cox 1. Should I use it instead? What is the meaning of 1 (is it just ln of HR)? Thank you! Sergey -----Message d'origine----- De : owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] De la part de Steve Samuels Envoyé : mardi 28 février 2012 23:48 À : statalist@hsphsun2.harvard.edu Objet : Re: st: Power estimation in Cox ph model Welcome to Statalist, Sergey! See the entry for -stpower- in the Stata Survival manual. By the way, the correct spelling of our favorite program is "Stata", not "STATA". In English, words all capitalized are "acronyms", meaning that their letters stood for words in the original, sometimes abandoned, names. So, for example, "SAS" and "SPSS" are acronyms, because their original names were "Statistical Analysis System" and "Statistical Package for the Social Sciences". "Stata" is a name the company made up and its letters don't stand for anything else (see the end of the Statalist FAQ), so the only capitalized letter is the initial "S", just as in our names. Steve On Feb 28, 2012, at 5:44 AM, ZHIVIN Sergey wrote: we want to do several models ( 1-univariate model with number of death from cancer and exposure to different chemicals, 2-the previous model adjusted for exposure to internal radiation). I have done a literature review and now I know hazard ratios of such exposures in previous studies. The problem I don't know how to construct my STATA formula for power. Could you help me with this? Number of deaths and study sample are known. Have a good day, Sergey Zhivin * * For searches and help try: * http://www.stata.com/help.cgi?search * * 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/

**References**:**st: Power estimation in Cox ph model***From:*"ZHIVIN Sergey" <Sergey.ZHIVIN@irsn.fr>

**Re: st: Power estimation in Cox ph model***From:*Steve Samuels <sjsamuels@gmail.com>

**RE: st: Power estimation in Cox ph model***From:*"ZHIVIN Sergey" <Sergey.ZHIVIN@irsn.fr>

- Prev by Date:
**st: RE: Hessian trigger "too many literals" error with "ml model"** - Next by Date:
**Re: st: Propensity Score Matching** - Previous by thread:
**RE: st: Power estimation in Cox ph model** - Next by thread:
**st: Weighted 2SLS** - Index(es):