From: Peter Dalgaard <P.Dalgaard_at_biostat.ku.dk>

Date: Fri, 20 Jun 2008 13:46:24 +0200

Date: Fri, 20 Jun 2008 13:46:24 +0200

Michael Pearmain wrote:

> Thanks for the reply Peter,

*>
**>
**>> I did just see that i had put the first error message,(agreed rather an
**>> obvious error) in and not the second i received
**>>
**>> Warning message:
**>> In asin(sqrt(yF[i])) : NaNs produced
**>>
*

Ah. Notwithstanding other remarks, that one would be due to

yT<-x[,"True"]/colSums(x)

The divisor here is a vector of length two!! So every 2nd time you are dividing with the wrong thing and in one case that gets you a relative frequency bigger than 1...

>> The reason i'm looking at this is advert exposure True and False.

*>>
**>> I'm inspecting age to asses weather or not to weight data in order to
**>> normalise groups for later questions,
**>> The questions that i am looking at later on are not scale based questions
**>> so i cannot perform t-tests on these, so i thought the only viable way was
**>> to look at z-tests for proportions to check for post-hoc differences
**>>
**>> Any advise on other methods would be gratefully taken
**>>
**>>
*

I still don't see where the notion of power needs to come in. Shouldn't
the effect size for "The questions that I am looking at later" enter
somehow? If you just want to know whether there is age-dependent
exposure, how about

chisq.test(SURVEY$n_0,exposed)

(or wilcox.test if you want something that takes the group ordering into account, or prop.trend.test). If, for better or worse, you want to test single rows, you could use

chisq.test(SURVEY$n_0 == "35-44",exposed)

etc.

*>>
*

>> On Fri, Jun 20, 2008 at 11:14 AM, Peter Dalgaard <P.Dalgaard@biostat.ku.dk>

*>> wrote:
**>>
**>>
**>>> Michael Pearmain wrote:
**>>>
**>>>> I'm having trouble creating a looping variable and i can't see wher ethe
**>>>> problem arises from any hep gratfully appreciated
**>>>>
**>>>> First create a table
**>>>>
**>>>> x<-table(SURVEY$n_0,exposed)
**>>>>
**>>>>
**>>>>> x
**>>>>>
**>>>>>
**>>>> exposed
**>>>> False True
**>>>> Under 16 24 1
**>>>> 16-19 68 9
**>>>> 20-24 190 37
**>>>> 25-34 555 204
**>>>> 35-44 330 87
**>>>> 45-54 198 65
**>>>> 55-64 67 35
**>>>> 65+ 10 8
**>>>>
**>>>> Now ectors to store counts and column proportions
**>>>>
**>>>>
**>>>>
**>>>>> xT<-x[,"True"]
**>>>>> xF<-x[,"False"]
**>>>>> yT<-x[,"True"]/colSums(x)
**>>>>> yF<-x[,"False"]/colSums(x)
**>>>>>
**>>>>>
**>>>> check length for dynamic looping
**>>>>
**>>>>
**>>>>> length(yT)
**>>>>>
**>>>>>
**>>>> [1] 8
**>>>>
**>>>> now create loop
**>>>>
**>>>>
**>>>>> for(i in 1:length(yT)){
**>>>>>
**>>>>>
**>>>> +
**>>>>
**>>> pwr.2p2n.test(2*(asin(sqrt(yT[i]))-asin(sqrt(yF[i]))),n1=xT[i],n2=xF[i])
**>>>
**>>>> + }
**>>>> Error in pwr.2p2n.test(2 * (asin(sqrt(yT[i])) - asin(sqrt(yF[i]))), n1 =
**>>>> xT[i], :
**>>>> number of observations in the first group must be at least 2
**>>>>
**>>>> this confuses me as if i enter the data as values the procedure works?
**>>>>
**>>>> Thanks in advance
**>>>>
**>>>>
**>>> Er, the first row "under 16" has a count of 1 in the "True" column and
**>>> it confuses you that you get an error saying that you need at least 2??
**>>>
**>>> But what looks _really_ confused is what you are trying to do in the
**>>> first place: The p's you are passing to pwr.2p2n are the empirical
**>>> relative frequencies of the individual age groups. This sort of reverses
**>>> cause and effect (presumably the exposure does not cause middle age) and
**>>> it is pretty odd to compare a particular row in a table with everything
**>>> else jumbled together but worse, it is post-hoc power calculation, which
**>>> is just a plain Bad Idea (as several people have pointed out before).
**>>>
**>>> --
**>>> O__ ---- Peter Dalgaard �ster Farimagsgade 5, Entr.B
**>>> c/ /'_ --- Dept. of Biostatistics PO Box 2099, 1014 Cph. K
**>>> (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918
**>>> ~~~~~~~~~~ - (p.dalgaard_at_biostat.ku.dk) FAX: (+45) 35327907
**>>>
**>>>
**>>>
**>>>
**>> --
**>> Michael Pearmain
**>> Senior Statistical Analyst
**>>
**>>
**>> 1st Floor, 180 Great Portland St. London W1W 5QZ
**>> t +44 (0) 2032191684
**>> mpearmain_at_google.com
**>> mpearmain_at_doubleclick.com
**>>
**>>
**>> Doubleclick is a part of the Google group of companies
**>>
**>
**> [[alternative HTML version deleted]]
**>
**>
**> ------------------------------------------------------------------------
**>
**> ______________________________________________
**> R-help_at_r-project.org mailing list
**> https://stat.ethz.ch/mailman/listinfo/r-help
**> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
**> and provide commented, minimal, self-contained, reproducible code.
**>
*

-- O__ ---- Peter Dalgaard Øster Farimagsgade 5, Entr.B c/ /'_ --- Dept. of Biostatistics PO Box 2099, 1014 Cph. K (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918 ~~~~~~~~~~ - (p.dalgaard_at_biostat.ku.dk) FAX: (+45) 35327907 ______________________________________________ R-help_at_r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.Received on Fri 20 Jun 2008 - 14:04:44 GMT

Archive maintained by Robert King, hosted by
the discipline of
statistics at the
University of Newcastle,
Australia.

Archive generated by hypermail 2.2.0, at Fri 20 Jun 2008 - 15:00:56 GMT.

*
Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-help.
Please read the posting
guide before posting to the list.
*