Chi_square

[Co Ar]

Dear all,
Am using this command:

bys Shock: ta Shock Treatment, col Chi2

However, no chi test gets registered. What can I do?

[Marcos Almeida]

If I understood your query, when you use ‘by shock’ beforehand and wish to tabulate “by shock” as well, there won’t be more than one level of shocks, hence no chi2.

[Co Ar]

There are many shocks, I just coded the variable Shock. But I would like to get the chi square for every category of shock and every category og "Treatment" variable. I have also triedn this command and nothing comes out:

tab Shock Treatment if Shock==1, col chi2

All these commands produce no chi square test

[Marcos Almeida]

I believe the explanation in #2 still applies, since you provided a similar example.

That said, I gather this toy example will be helpful to clarify your doubt:

. sysuse auto
(1978 Automobile Data)

. gen highmpg = mpg > 27 & !missing(mpg)

. tab foreign highmpg, chi2

           |        highmpg
  Car type |         0          1 |     Total
-----------+----------------------+----------
  Domestic |        47          5 |        52
   Foreign |        16          6 |        22
-----------+----------------------+----------
     Total |        63         11 |        74

          Pearson chi2(1) =   3.8087   Pr = 0.051

. tab foreign highmpg if foreign ==1, chi2

           |        highmpg
  Car type |         0          1 |     Total
-----------+----------------------+----------
   Foreign |        16          6 |        22
-----------+----------------------+----------
     Total |        16          6 |        22


. by foreign, sort: tab foreign highmpg, chi2

------------------------------------------------------------------------------------------------------------
-> foreign = Domestic

           |        highmpg
  Car type |         0          1 |     Total
-----------+----------------------+----------
  Domestic |        47          5 |        52
-----------+----------------------+----------
     Total |        47          5 |        52


------------------------------------------------------------------------------------------------------------
-> foreign = Foreign

           |        highmpg
  Car type |         0          1 |     Total
-----------+----------------------+----------
   Foreign |        16          6 |        22
-----------+----------------------+----------
     Total |        16          6 |        22

As we see, as long as we get only one level of a categorical variable, it would be nonsense (or impossible, if you will) to apply a chi-squared test, which means measures of assocation betwen categorical variables. In the last examples, where the chi-squared test is not feasible, there is only one level for the categorical variable.

I hope to have helped.

[Co Ar]

I see the sense. I did not pay attention to the categories.
This is now understood