How to count the number of relevant variables present for an individual

[Thea Black]

I have a multiple cross-sectional dataset, where I am trying to create the following index:

Dit = (Sum all dit until k)/Kt

In other words, sum the values for all d variables and then divide by the number d variables for that year.

In my dataset, i have 4 variables that i need to look at but for some years only three or two are present so to calculate my index I don't want to just include individuals that had all four. I don't want them to be treated as missing. Instead, i want the aggregated variables to be divided by say 3 instead of 4 if only 3 variables are present. How do i go about creating this Kt?

My four variables are: racmar, racpeers, racpres and racseg. They are byte variables (not string) and are non binary.

[Carlo Lazzaro]

Thea:
do you mean something along the following lines?

. set obs 3
number of observations (_N) was 0, now 3

. g id=_n

. g A=runiform() in 1/2

. g B=runiform() in 2/3

. g C=runiform() if _n!=2

. g D=runiform()

. egen Overall=rowtotal(A-D)

. egen Ratio=rownonmiss(A-D)

. egen Denominator=rownonmiss(A-D)

. g Ratio=Overall/Denominator

. list

     +---------------------------------------------------------------------------------+
     | id          A          B          C          D    Overall   Denomi~r      Ratio |
     |---------------------------------------------------------------------------------|
  1. |  1   .3488717          .   .8689333   .0711051    1.28891          3   .4296367 |
  2. |  2   .2668857   .1366463          .    .323368      .7269          3      .2423 |
  3. |  3          .   .0285569   .3508549   .5551032   .9345149          3    .311505 |
     +---------------------------------------------------------------------------------+

.