converting float variable to integer

[Tess Gecha]

Hi,

I'd like to convert a float variable to an integer. I've tried "recast int var" and it did not work. Can someone please help?

[Clyde Schechter]

"Did not work" is not helpful. There is only one way a command can "work," but a large number of ways it can fail. Did you get an error message? If so, what was it? Did Stata crash? Did it run but produce results different from what you expected? (If so, how did the results differ from what you expected?) Usually the best way to pose a question like this is to show, by direct copy and paste from the Results window or your log file, the exact command you typed and the exact output/results that Stata gave you. It is also usually help to provide example data the exhibits the problem.

[Rich Goldstein]

in addition to what Clyde Schechter said, all of which I agree with, you might want to try

compress var

as it sometimes gives more informative messages; see

help compress

[Leonardo Guizzetti]

recast int varname

Will issue a warning when values cannot be converted from float/double to int. If there are no issues, then nothing is displayed. -compress- would also automatically recast from float/long to byte/int/long as long as no issues such as a decimals are detected, which is Rich's suggestion.

Since it is likely you do not have only integer values, you can explore them with something like the following, where you replace x with your own variable name.

list x if mod(x,1) != 1
tab x if mod(x,1) != 1

[Tess Gecha]

After running compress numsexpart, it is still stored as a float variable. I've copy and pasted the output below.


compress numsexpart
(0 bytes saved)

. describe numsexpart

storage display value
variable name type format label variable label
-------------------------------------------------------------------------------------------------
numsexpart float %3.0f How many sexual partners have you had




I also ran recast int numsexpart and again, it still saved as a float variable.


recast int numsexpart
numsexpart: 5 values would be changed; not changed

. describe numsexpart

storage display value
variable name type format label variable label
-------------------------------------------------------------------------------------------------
numsexpart float %3.0f How many sexual partners have you had



It also appears that they are all integer values?


tab numsexpart if mod(numsexpart,1) != 1

How many |
sexual |
partners |
have you |
had | Freq. Percent Cum.
------------+-----------------------------------
0 | 5 4.76 4.76
1 | 10 9.52 14.29
2 | 8 7.62 21.90
3 | 3 2.86 24.76
4 | 4 3.81 28.57
5 | 9 8.57 37.14
6 | 5 4.76 41.90
7 | 5 4.76 46.67
8 | 5 4.76 51.43
10 | 12 11.43 62.86
11 | 2 1.90 64.76
12 | 2 1.90 66.67
15 | 3 2.86 69.52
16 | 1 0.95 70.48
17 | 1 0.95 71.43
18 | 1 0.95 72.38
20 | 7 6.67 79.05
22 | 1 0.95 80.00
23 | 2 1.90 81.90
24 | 1 0.95 82.86
25 | 3 2.86 85.71
30 | 4 3.81 89.52
40 | 1 0.95 90.48
50 | 1 0.95 91.43
52 | 1 0.95 92.38
70 | 1 0.95 93.33
75 | 2 1.90 95.24
100 | 3 2.86 98.10
250 | 1 0.95 99.05
300 | 1 0.95 100.00
------------+-----------------------------------
Total | 105 100.00

.

[Tess Gecha]

This is a better view of the tab numsexpart if mod(numsexpart,1) != 1 command

[Maarten Buis]

There is a difference between what you see on screen and what is stored inside the computer. What counts is what is stored inside the computer, not what is shown on the screen. The display format %3.0f of your variable hides all fractions. Given the command you executed those numbers you see have to be non-integers, it is just the display format that hides the fractional part. So type in Stata

format numsexpart %9.3f

and repeat that tab command again. This won't solve your problem, but allows you to diagnose what the problem might be.

[Leonardo Guizzetti]

My apologies, I had a typo in #4 that may have led you a bit astray. The if qualifier should have been

mod(x,1) != 0

Which means, the remainder after repeatedly dividing out by 1 is not equal to zero (i.e., a fraction).

The following result you showed in #5 shows you do have 5 non-integer values which you can try to resolve.

. recast int numsexpart
numsexpart: 5 values would be changed; not changed

Perhaps a cleaner way to do the same thing:

clear
input float x
2.4
4.8
  3
  .
end

list if int(x) != x

Result

     +-----+
     |   x |
     |-----|
  1. | 2.4 |
  2. | 4.8 |
     +-----+