HI everyone
I think we can close this one now. This is a case where I am reminded
of the person who said (Clinton?) "The American people have spoken but
it might take a few days to work out what they said" and that won't be
today. If anyone is interested in the answers I might have got my head
around it by about Friday!
thanks everyone
Andy
On 10 May 2016 at 13:25, Richard Goldstein <[log in to unmask]> wrote:
> Your welcome - I, at least, would appreciate it if you would "close" the
> thread by providing the response to your issue
>
> On 5/10/16 8:20 AM, Andrew Salmon wrote:
>> Thanks Richard, didn't think of that but Roger as it happens has got
>> back to me already, plus some others with very helpful stuff
>>
>> On 10 May 2016 at 13:00, Richard Goldstein <[log in to unmask]> wrote:
>>> I suggest posting this on Statalist (http://www.statalist.org/); somersd
>>> is a user-written package and the author, Roger Newson, appears to
>>> follow the Statalist discussions - you could also email him directly -
>>> his email address is at the bottom of the help file
>>>
>>> Rich
>>>
>>> On 5/10/16 6:00 AM, Andrew Salmon wrote:
>>>> Hi all
>>>> am just investigating the finer details of the Mann Whitney test and
>>>> trying to calculate confidence intervals for the porder statistic. I
>>>> have the following dummy data with var2 as the grouping variable
>>>>
>>>> using STATA's ranksum with porder option gives me a value of 0.594
>>>> (for group = 0 over group = 1). Doing the same calculation using the
>>>> somersd package with transf(c) option gives me a confidence interval.
>>>> ( I used vreverse to generate rgroup)
>>>>
>>>> | Jackknife
>>>> rgroup | Coef. Std. Err. z P>|z| [95% Conf. Interval]
>>>> -------------+----------------------------------------------------------------
>>>> var1 | .59375 .1594069 3.72 0.000 .2813182 .9061818
>>>> -----------------------------------------------------------------------------
>>>> but with a p value that seems rather low and doesn't seem to make
>>>> sense with the confidence interval, since I would have thought the
>>>> null of probability/proportion would be 0.5
>>>>
>>>> Any thoughts?
>>>>
>>>> thanks
>>>> Andy
>>>>
>>>> var1 var2
>>>> 1 0
>>>> 5 0
>>>> 9 0
>>>> 14 0
>>>> 21 0
>>>> 29 0
>>>> 39 0
>>>> 49 0
>>>> 3 1
>>>> 4.5 1
>>>> 7 1
>>>> 12 1
>>>> 19 1
>>>> 20.5 1
>>>> 26 1
>>>> 30 1
>>
>> You may leave the list at any time by sending the command
>>
>> SIGNOFF allstat
>>
>> to [log in to unmask], leaving the subject line blank.
>>
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|