creating a new variable out of *one standard deviation above* its mean

[ishak celik]

Hi, I am trying to create a new variable out of one standard deviation above its mean. I used some codes, it did not help. Generating a new variable capturing for example one standard deviation above mean math grades? Thank you!

[Nick Cox]

This may help:

. sysuse auto, clear
(1978 automobile data)

. su mpg

    Variable |        Obs        Mean    Std. dev.       Min        Max
-------------+---------------------------------------------------------
         mpg |         74     21.2973    5.785503         12         41

. gen himpg = mpg > (r(mean) + r(sd)) if mpg < .

. tab himpg

      himpg |      Freq.     Percent        Cum.
------------+-----------------------------------
          0 |         63       85.14       85.14
          1 |         11       14.86      100.00
------------+-----------------------------------
      Total |         74      100.00

[Maarten Buis]

Your question is not very clear. For example the sentence "I used some codes, it did not help." contains two really big problems: First, you did not tell what code you used. We cannot fix unknown code. Second, "it did not help" could mean many things: it gave an error message, it gave results but not the ones i want, it did nothing, etc. etc. So a good question will tell us:

• what your data looks like by presenting an example (see help dataex),
• what you want to achieve. In your case it is unclear what your new variable is supposed to look like. Is it the value when it exceeds the cutoff and missing otherwise, is it supposed to be an indicator/dummy variable, or something else
• what you have tried. Give us the exact code
• why the results are not what you want. Don't assume that is obvious, very often it is not.

[ishak celik]

I have a variable like this.


Variable | Obs Mean Std. Dev. Min Max
-------------+---------------------------------------------------------
var | 27,297 1.131764 .3169751 1 4

I used above codes and found a result that does not give meaning

gen hi_var = var > (r(mean) + r(sd)) if var< .


. tab hi_var


hi_var | Freq. Percent Cum.
------------+-----------------------------------
0 | 27,297 100.00 100.00
------------+-----------------------------------
Total | 27,297 100.00


What can be the problem?

[ishak celik]

Your question is not very clear. For example the sentence "I used some codes, it did not help." contains two really big problems: First, you did not tell what code you used. We cannot fix unknown code. Second, "it did not help" could mean many things: it gave an error message, it gave results but not the ones i want, it did nothing, etc. etc. So a good question will tell us:

• what your data looks like by presenting an example (see help dataex),
• what you want to achieve. In your case it is unclear what your new variable is supposed to look like. Is it the value when it exceeds the cutoff and missing otherwise, is it supposed to be an indicator/dummy variable, or something else
• what you have tried. Give us the exact code
• why the results are not what you want. Don't assume that is obvious, very often it is not.

Hi, I would like to create a dummy variable "one standard deviation above the mean". I checked forum before and tried some codes that I saw. here is the variable for example:

.summ var1

Variable | Obs Mean Std. Dev. Min Max
-------------+---------------------------------------------------------
var1 | 26,004 1.21038 .4415799 1 5

when I use this code : gen hi_var1 = var1 > (r(mean) + r(sd)) if var1 < .
I have a result that looks like this:

. tab hi_var1

hi_var1 | Freq. Percent Cum.
------------+-----------------------------------
0 | 26,004 100.00 100.00
------------+-----------------------------------
Total | 26,004 100.00
It seems that I could not have a dummy variable out of this code. I don't know why is that? can you guys suggest a solution for this?

[Maarten Buis]

Is that a continuous variable or a ordered variable with values 1,2,3,4, or 5 (maybe a Likert scale)? If the latter is the case, then the whole mean + standard deviation as a cut-off makes little sense. Especially since then there is often a more substantively interesting cut-offs easily available.

The code you used has to happen immediately after the sum command, as it is the sum command that temporarily creates the r(mean) and r(sd).

[ishak celik]

Ok, 1. the variable is an averaged index of ordinal level variables. (5 questions measuring depressive symptoms with 5 points Likert like frequency points). 2. when I run it right afterthe sum command it gives a result. Thanks.

[Amelie Nickel]

Dear Nick Cox

I have an similar issue like the post #1 https://www.statalist.org/forums/for...an#post1651574

I want to generate a new variable that categorizes a numeric variable into
“average” (mean)
"high” (1.5 standard deviations above the mean)
“low” (1.5 standard deviations below the mean)

It would be great help, if you might provide the stata code for that using your example (below).

Thank you very much.

This may help:

. sysuse auto, clear
(1978 automobile data)

. su mpg

Variable | Obs Mean Std. dev. Min Max
-------------+---------------------------------------------------------
mpg | 74 21.2973 5.785503 12 41

. gen himpg = mpg > (r(mean) + r(sd)) if mpg < .

. tab himpg

himpg | Freq. Percent Cum.
------------+-----------------------------------
0 | 63 85.14 85.14
1 | 11 14.86 100.00
------------+-----------------------------------
Total | 74 100.00

[Nick Cox]

. sysuse auto, clear
(1978 automobile data)

. su mpg

    Variable |        Obs        Mean    Std. dev.       Min        Max
-------------+---------------------------------------------------------
         mpg |         74     21.2973    5.785503         12         41

. gen wanted = cond(mpg < r(mean) - 1.5 * r(sd), 1, cond(mpg < r(mean) + 1.5 * r(sd), 2, 3)) if mpg <   .

. tab mpg wanted, missing

   Mileage |              wanted
     (mpg) |         1          2          3 |     Total
-----------+---------------------------------+----------
        12 |         2          0          0 |         2 
        14 |         0          6          0 |         6 
        15 |         0          2          0 |         2 
        16 |         0          4          0 |         4 
        17 |         0          4          0 |         4 
        18 |         0          9          0 |         9 
        19 |         0          8          0 |         8 
        20 |         0          3          0 |         3 
        21 |         0          5          0 |         5 
        22 |         0          5          0 |         5 
        23 |         0          3          0 |         3 
        24 |         0          4          0 |         4 
        25 |         0          5          0 |         5 
        26 |         0          3          0 |         3 
        28 |         0          3          0 |         3 
        29 |         0          1          0 |         1 
        30 |         0          0          2 |         2 
        31 |         0          0          1 |         1 
        34 |         0          0          1 |         1 
        35 |         0          0          2 |         2 
        41 |         0          0          1 |         1 
-----------+---------------------------------+----------
     Total |         2         65          7 |        74 

. 

.  scatter wanted mpg, ms(none) mlab(wanted) mlabpos(0) yla( 1 2 3) 
.

[Amelie Nickel]

Thank you very much!

[Nick Cox]

However, nothing special ever happens 1.5 SD away from the mean so far as I know, so you're throwing away information....