Roger is correct. Among other things, -koopmani- works with a scalar called m.
As documented in
SJ-6-2 dm0021 . Stata tip 31: Scalar or variable? Problem of ambiguous names
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . G. I. Kolev
Q2/06 SJ 6(2):279--280 (no commands)
tips for avoiding abbreviation conflicts with variables
when naming scalars
-- and in the manual entries it refers to -- scalars and variables share the same name space. Thus given a reference to "m", Stata tries first to make sense of it as a reference to a variable.
Roger's example points up directly and indirectly how this works with the auto dataset. Stata, on behalf of -koopmani-, sees "m" as a possible abbreviation for -make- and -mpg-, but complains about ambiguity. Drop -make- and Stata does not complain but it still produces nonsense because it interprets m as a reference to -mpg-, meaning -mpg[1]- which only by a freakish coincidence will contain the same number as the scalar m.
The main advice: don't do that! Using temporary names is much better practice.
As -koopmani- was published in the STB, we'll encourage the author to fix this (and bring the program presentation and documentation up-to-date too).
Nick
[email protected]
(also wearing an Editor SJ hat)
Roger Harbord
-koopmani- seems to have a programming issue as well as that labelling issue:
. sysuse auto, clear
(1978 Automobile Data)
. koopmani 36 40 16 80
m ambiguous abbreviation
r(111);
. drop make
. koopmani 36 40 16 80
| Event | Proportion
| Yes No | Total Yes
-----------------+------------------------+----------------------
Group1 | 36 -14 | 22 1.6364
Group2 | 16 64 | 80 0.2000
-----------------+------------------------+----------------------
Total | 52 50 | 102 0.5098
| |
| Point estimate | [95% Conf. Interval]
-----------------|------------------------+----------------------
Odds Ratio | 8.181818 | 7.313014 12.16772
-----------------------------------------------------------------
. drop _all
. koopmani 36 40 16 80
| Event | Proportion
| Yes No | Total Yes
-----------------+------------------------+----------------------
Group1 | 36 4 | 40 0.9000
Group2 | 16 64 | 80 0.2000
-----------------+------------------------+----------------------
Total | 52 68 | 120 0.4333
| |
| Point estimate | [95% Conf. Interval]
-----------------|------------------------+----------------------
Odds Ratio | 4.5 | 2.939633 7.152252
-----------------------------------------------------------------
I suspect koopmani.ado would benefit from some -tempvar- or tempname-
statements.
Maarten Buis
> I assume you refer to this -koopmani-:
> STB-58 sg154 . Conf. int. for ratio of two binomial prop. by Koopman's method (help koopman if installed) .. . . . . . . . . . . . . . D. Wang
> 11/00 pp.16--19; STB Reprints Vol 10, pp.244--247
>
> From the helpfile of -koopmani-: "koopman computes the confidence
> intervals for the ratio of two binomial proportions", so it is a risk
> ratio not an odds ratio.
Tiago Pereira
Working with -koopmani-, one types:
. koopmani 36 40 16 80
and gets:
| Event | Proportion
| Yes No | Total Yes
-----------------+------------------------+----------------------
Group1 | 36 4 | 40 0.9000
Group2 | 16 64 | 80 0.2000
-----------------+------------------------+----------------------
Total | 52 68 | 120 0.4333
| |
| Point estimate | [95% Conf. Interval]
-----------------|------------------------+----------------------
Odds Ratio | 4.5 | 2.939633 7.152252
-----------------------------------------------------------------
That estimate of 4.5 is actually the risk ratio, isn´t it? Or am I doing a big confusion?
. cci 36 4 16 64
*
* 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/