Descriptive Statistics by Segment

[Asteris Dougalis]

Hi all,

I have ratings data (ratings are 0-5) and I am trying to see what what % of users never rate below 3. This is the code that I have so far. Note that user_ratingscount is the the number of ratings left by a user at the time of each rating.

    sort userid 
    bysort userid: gen rating_below_3 = 1 if rating < 3
    replace rating_below_3 = 0 if rating <= 3 
    egen max = max(user_ratingscount), by(userid)
    egen sum3 = total(rating_below_3), by(userid)
    bysort userid: gen never_rated_below_3 = 1 if max == sum
    bysort userid: replace never_rated_below_3 = 0 if max != sum 
    duplicates drop userid, force 
    tab never_rated_below_3 

I am running into trouble when a user has rated below a 3 but has never rated exactly a 3. The code above comes out as 1 in that case even though it should still report a 0 for the never_rated_below_3 variable. Here is a data example.

* Example generated by -dataex-. To install: ssc install dataex
clear
input double(userid rating user_ratingscount) float(rating_below_3 max sum3 never_rated_below_3)
 2 4.5  4 . 12 0 0
 2   3  1 0 12 0 0
 2   3  9 0 12 0 0
 2   3  8 0 12 0 0
 2   4 11 . 12 0 0
 2 3.5  3 . 12 0 0
 2   4 10 . 12 0 0
 2 4.5  7 . 12 0 0
 2   5 12 . 12 0 0
 2   4  2 . 12 0 0
 2   4  6 . 12 0 0
 2   4  5 . 12 0 0
 3   5  1 .  1 0 0
 4 3.5  9 . 13 0 0
 4 3.5 11 . 13 0 0
 4   5  2 . 13 0 0
 4 3.5  3 . 13 0 0
 4   5  5 . 13 0 0
 4   5 12 . 13 0 0
 4   5 10 . 13 0 0
 4 3.5  4 . 13 0 0
 4 4.5  7 . 13 0 0
 4   5 13 . 13 0 0
 4   5  8 . 13 0 0
 4   5  1 . 13 0 0
 4   5  6 . 13 0 0
 5 2.5 32 0 38 0 0
 5   3  6 0 38 0 0
 5   5 20 . 38 0 0
 5   5 37 . 38 0 0
 5   5 35 . 38 0 0
 5   5 33 . 38 0 0
 5   5 25 . 38 0 0
 5   5  8 . 38 0 0
 5 4.5 18 . 38 0 0
 5   5 28 . 38 0 0
 5   5 13 . 38 0 0
 5   5 17 . 38 0 0
 5   5 31 . 38 0 0
 5 4.5 12 . 38 0 0
 5   5 34 . 38 0 0
 5   5  2 . 38 0 0
 5   5  7 . 38 0 0
 5   5 21 . 38 0 0
 5   4 29 . 38 0 0
 5   5  5 . 38 0 0
 5   5 30 . 38 0 0
 5   5 22 . 38 0 0
 5   5  4 . 38 0 0
 5   5 15 . 38 0 0
 5 4.5  1 . 38 0 0
 5   5 10 . 38 0 0
 5   5 19 . 38 0 0
 5   5 23 . 38 0 0
 5   4 27 . 38 0 0
 5   5 36 . 38 0 0
 5   5  3 . 38 0 0
 5   5 38 . 38 0 0
 5   4 16 . 38 0 0
 5   5 11 . 38 0 0
 5   5 14 . 38 0 0
 5   5  9 . 38 0 0
 5   5 24 . 38 0 0
 5   5 26 . 38 0 0
 6   5  8 .  8 0 0
 6   5  5 .  8 0 0
 6   4  6 .  8 0 0
 6   5  2 .  8 0 0
 6   4  3 .  8 0 0
 6   5  4 .  8 0 0
 6   3  7 0  8 0 0
 6   3  1 0  8 0 0
 7   5  1 .  1 0 0
 8   3  2 0  2 0 0
 8   4  1 .  2 0 0
 9 3.5  2 .  2 0 0
 9   2  1 0  2 0 0
10 4.5  6 .  8 0 0
10   5  5 .  8 0 0
10   3  2 0  8 0 0
10   5  8 .  8 0 0
10 2.5  7 0  8 0 0
10   5  1 .  8 0 0
10   3  4 0  8 0 0
10 4.5  3 .  8 0 0
11   5 13 . 13 0 0
11   5 12 . 13 0 0
11 4.5  5 . 13 0 0
11   5  6 . 13 0 0
11   5  2 . 13 0 0
11   5 11 . 13 0 0
11   4 10 . 13 0 0
11   4  4 . 13 0 0
11   5  8 . 13 0 0
11   5  1 . 13 0 0
11 4.5  3 . 13 0 0
11   5  7 . 13 0 0
11 4.5  9 . 13 0 0
12   4  7 . 14 0 0
12 3.5  6 . 14 0 0
end

Does anyone have a more efficient way to complete the task or a possible edit to the code above? Thank you in advance for all the help.

[Carlo Lazzaro]

Asteris:
are you sure that the second line of your code should not be:

replace rating_below_3 = 0 if rating >= 3

[Asteris Dougalis]

Hi. Thanks for pointing that typo out. Even after correcting it, though, the error I described in #1 persists. Do you have any insights on a more efficient way to complete the task?

Best,

Asteris

[Carlo Lazzaro]

Asteris:
you may want to try:

 bysort userid: gen wanted=1 if rating<3


. collapse (count) wanted, by(userid) ///remember to save a copy of your dataset before -collapse-///

. tab wanted

    (count) |
     wanted |      Freq.     Percent        Cum.
------------+-----------------------------------
          0 |          8       72.73       72.73
          1 |          3       27.27      100.00
------------+-----------------------------------
      Total |         11      100.00

.

[Asteris Dougalis]

That worked! Thanks.