Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Re: RE: Re: RE: one-sample chi square test


From   "Wendell Joice" <wjoice@erols.com>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: Re: RE: Re: RE: one-sample chi square test
Date   Sat, 1 Feb 2003 14:53:14 -0500

Nick,
Thanks so much for your advice. Your second (sequel) approach mentioned at
the bottom of this message works fine. For folks like me, this assistance is
a KEY reason why I chose and stick with Stata.

My attempts to use your first approach, immediately below, yielded an error
message as you can see.
chitest  rep78, count
      count invalid
      r(198);

But I have been successful using your second approach; it took me a while
since this is my first experience with macros and multi-step commands
like -forvalues-. But this was bonus learning for me.
As you suggested, I added a couple of things and now have a do file [listed
below]  (with arguments) that has been working fine, especially since I have
no data with '0' observed values.

DO FILE
.do chitest SEX "& racewbh==3 & csample==1"

   .forval i = 0/5 {
        2. qui count if  `1'== `i' `2'
        3. if r(N)==0 {
        4. continue
        5. }
        6. local counts "`counts' `r(N)'"
        7. }

   . di "`counts'
       950 767

   . chitesti `counts'

      Chi-square test:
          observed frequencies from keyboard
          expected frequencies equal

               Pearson chi2(1) =  19.5044   Pr =  0.000
      likelihood-ratio chi2(1) =  19.5415   Pr =  0.000

                                         residuals
            observed    expected     classic     Pearson
        1.       950     858.500      91.500       3.123
        2.       767     858.500     -91.500      -3.123

   . macro drop _counts

   . end of do-file


----- Original Message -----
From: "Nick Cox" <n.j.cox@durham.ac.uk>
To: <statalist@hsphsun2.harvard.edu>
Sent: Thursday, January 30, 2003 5:31 AM
Subject: st: RE: Re: RE: one-sample chi square test


> Wendell Joice
>
> > -chitesti- works fine; I need to tabulate the variables of
> > interest and then key in the frequencies to -chitesti-.
> >
> > Perchance, however, is there a command (other than and not
> > as elaborate
> > as -collapse-), that will produce a variable B containing
> > the frequencies of
> > the values in variable A? That way, I can place variable B directly
> > into -chitest- .
> >
> > Unless, of course, you know of a one-sample test that will
> > carry out the
> > entire process of  calculating the frequencies of the
> > values of a specified
> > categorical variable and then provide a one-sample
> > chi-square over those
> > frequencies? (Analogous to -tabulate- for two sample chi square)
> >
>
> -chitest- is able to do this for you. Here is
> a silly example for the simplest case of equal
> expected frequencies.
>
> . chitest rep78, count
>
> Chi-square test:
>     observed frequencies from rep78
>     expected frequencies equal
>
>          Pearson chi2(4) =  33.3913   Pr =  0.000
> likelihood-ratio chi2(4) =  34.7183   Pr =  0.000
>
>                                    residuals
>      +-----------------------------------------+
>      | observed   expected   classic   Pearson |
>      |-----------------------------------------|
>   1. |        2     13.800   -11.800    -3.176 |
>   2. |        8     13.800    -5.800    -1.561 |
>   3. |       30     13.800    16.200     4.361 |
>   4. |       18     13.800     4.200     1.131 |
>   5. |       11     13.800    -2.800    -0.754 |
>      +-----------------------------------------+
>
> What -chitest- can't do is count values which
> could be there in principle but which in
> practice don't occur (cells with observed 0,
> in short).
>
> I suspect there are other problems, but at worst
> you need to type a few numbers at -chitesti-.
>
> If this doesn't solve the problem, please give
> a clear example of the data structure you have in
> mind.
>
> Nick

SECOND APPROACH

As a sequel to my previous reply, here is one technique
illustrated:

. forval i = 1/5 {
  2. qui count if rep78 == `i'
  3. local counts "`counts' `r(N)'"
  4. }

. di "`counts'
 2 8 30 18 11

. chitesti `counts'

One advantage of this technique is that
will work fine when observed frequencies
are 0. You just have to use -forvalues-
or -foreach- to cycle over the possible
values of a variable relevant to your
test.

This could also be automated in a program.

Nick *




*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index