Calculting Difference in Proportion between two outcomes across a variable

[Frank Peter]

Dear All,


Calculting Difference in Proportion between two outcomes (out1 and out2) across a variable (agegrp:
Thanks a lot

. webuse bdesop, clear

. gen out1 = cases != 0

. gen out2 = cases > 5

. tab agegrp out1, row

+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+

| out1
agegrp | 0 1 | Total
-----------+----------------------+----------
25-34 | 28 2 | 30
| 93.33 6.67 | 100.00
-----------+----------------------+----------
35-44 | 20 10 | 30
| 66.67 33.33 | 100.00
-----------+----------------------+----------
45-54 | 6 26 | 32
| 18.75 81.25 | 100.00
-----------+----------------------+----------
55-64 | 0 32 | 32
| 0.00 100.00 | 100.00
-----------+----------------------+----------
65-74 | 2 28 | 30
| 6.67 93.33 | 100.00
-----------+----------------------+----------
75+ | 2 20 | 22
| 9.09 90.91 | 100.00
-----------+----------------------+----------
Total | 58 118 | 176
| 32.95 67.05 | 100.00

. tab agegrp out2, row

+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+

| out2
agegrp | 0 1 | Total
-----------+----------------------+----------
25-34 | 30 0 | 30
| 100.00 0.00 | 100.00
-----------+----------------------+----------
35-44 | 30 0 | 30
| 100.00 0.00 | 100.00
-----------+----------------------+----------
45-54 | 28 4 | 32
| 87.50 12.50 | 100.00
-----------+----------------------+----------
55-64 | 22 10 | 32
| 68.75 31.25 | 100.00
-----------+----------------------+----------
65-74 | 26 4 | 30
| 86.67 13.33 | 100.00
-----------+----------------------+----------
75+ | 22 0 | 22
| 100.00 0.00 | 100.00
-----------+----------------------+----------
Total | 158 18 | 176
| 89.77 10.23 | 100.00

.


Frank

[Bruce Weaver]

Hi Frank. Something like this, perhaps?

* NOTE:  Your out1 and out2 variables are not mutually exclusive!
tabulate out1 out2

* With 0/1 coding, the mean = proportion of 1s
bysort agegrp: egen p1 = mean(out1)
by agegrp: egen p2 = mean(out2)
generate pdiff = p1-p2
egen tag = tag(agegrp) // Tag one record per age group for listing
list agegrp p1 p2 pdiff if tag, clean noobs

Results from -list- command:

. list agegrp p1 p2 pdiff if tag, clean noobs

    agegrp         p1         p2      pdiff  
     25-34   .0666667          0   .0666667  
     35-44   .3333333          0   .3333333  
     45-54      .8125       .125      .6875  
     55-64          1      .3125      .6875  
     65-74   .9333333   .1333333         .8  
       75+   .9090909          0   .9090909

[Frank Peter]

Hi Bruce,

Thanks a lot. That is what I want. Please is it possible to calculate the 95% CI for the difference (Oliver Twist).


Frank