I replied to Paul O'Brien
> >
> > I have not been able to get the table layout right to do a
> > test for trend
> > (NPTREND) with the following data:
> >
> > WEIGHT EE20mcg EE30mcg
> > gained 48 24
> > stable 210 114
> > lost 38 18
> >
> > What I get is this:
> >
> > . tabi 48 24\210 114\38 18
> >
> > | col
> > row | 1 2 | Total
> > -----------+----------------------+----------
> > 1 | 48 24 | 72
> > 2 | 210 114 | 324
> > 3 | 38 18 | 56
> > -----------+----------------------+----------
> > Total | 296 156 | 452
> >
> > Pearson chi2(2) = 0.2483 Pr = 0.883
> >
> > . list
> >
> > row col pop
> > 1. 1 1 48
> > 2. 1 2 24
> > 3. 2 1 210
> > 4. 2 2 114
> > 5. 3 1 38
> > 6. 3 2 18
> >
> > . nptrend row, by(col)
> >
> > col score obs sum of ranks
> > 1 1 3 10.5
> > 2 2 3 10.5
> >
> > z = 0.00
> > Prob > |z| = 1.00
> >
> >
> >
> > The correct answer is:
> > One sided P = .2965
> > Two sided P = .593
> >
> > Where am I going wrong?
>
> The problem is that Stata does not
> know that it should look at your -pop- variable.
>
> Try
>
> . nptrend pop, by(row)
>
> Incidentally, a graph created with
>
> . ordplot row [w=pop] , by(col)
>
> supports the idea that these groups
> of responses are similar.
>
> (For -ordplot-, type -ssc desc ordplot-.)
>
No, I think my advice here on -nptrend-
is wrong: I followed your previous
path too readily. In fact, it
is not at all clear that this
is a problem for -nptrend- at
all. Your response appears
to be (ordered) categorical
and your covariate something
on two levels.
I think the graph still makes sense
and it suggest that there is no
relationship worth talking about.
Nick
[email protected]
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
