inconsistent result from contingency table (Analysis) and k-sample proportion test (Extras) on a 2X2 table

[eagle.e...@gmail.com]

For some reason I get a different result from contingency table in the Analysis menu and the comparison of proportion test in the Extras menu.

Using the birthwt dataset I get, what I think is the correct result from the proportion test (see below). However, the contingency function give a smaller calculated chi-square statistics (4.236 rather than 4.9237). Just to be sure I checked the calculations in excel (see pdf and xls file attached). I get the exact same result as the proportion test.

2-sample test for equality of proportions without continuity

correction

data:  table(group = birthwt$smoke, variable = birthwt$low) 

X-squared = 4.9237, df = 1, p-value = 0.02649

alternative hypothesis: two.sided 

95 percent confidence interval:

 0.01607177 0.29039121 

sample estimates:

   prop 1    prop 2 

0.7478261 0.5945946 

[Ian Fellows]

The analysis menu is giving the chi-squared test with yates continuity
correction.

chisq.test(table(birthwt$smoke,birthwt$low),correct=FALSE)

chisq.test(table(birthwt$smoke,birthwt$low))

I will consider making no correction the default.

ian

[Vincent]

I didn't realize a correction was applied. No correction as the default would be great. Or an option to turn it off/on in the contingency dialog.

Thanks Ian!