Using a chi square when comparing treatments

#1
Greetings,
I have the following data for two different medicines (one currently being used and one new). The data is number of treatments and number of pregnancies

Bemfola - AID (136, 29)
Gonal F - AID (262, 37)
Bemfola - AIH (95, 9)
Gonal F - AIH (150, 11)

The null hypothesis is that Bemfola produces the same amount of pregnancies as Gonal.

I personally compute this to the following for the AID group:
Observed = 29 (the actual amount of pregnancies)
Expected (37/262)*136 = 19 (the amount of pregnancies expected if Bemfola had the same percentage pregnancies)
Chi^2 = 4,99
p = 0,025

But when running it in SPSS I get an entirely different outcome:
Bemfola
Count = 29
Expected = 22,6
Gonal F
Count = 37
Expected = 43,4
Chi^2 = 3,356
p = 0,067


What am I doing wrong?
 
#2
Hello,

You can compute the expected cell count for a certain cell as row_total * column_total / overall_total. So if we only look within the AID group, the expected cell count in the bemfola + pregnant group is equal to 136 * 66 / 398 = 22.6. Here 136 is the total number of subjects in the Bemfola group, 66 is the total the number of pregnant subjects (29 + 37), and 398 is the total number of subjects (136 + 262).