The mean for different variables with different scales

[Sally Ahmed]

Hello,

I would like to know how can I get the average for some different variables which when they are combined they represent an indicator for one variable. What is the syntax to calculate the average to be able to reach the indicator which is calculated as the mean rank score of the four indices?

Thanks

[Nick Cox]

I don't see what you wanting precisely here. The recipe is presumably (a) to calculate rank scores and then (b) to take a row mean. (b) is easy to guess as best done with egen and just possibly (a) is too.

What is really important is to decide exactly what to do about ties and missing values.

[Sally Ahmed]

how can I calculate the rank score what is its syntax?

[Sally Ahmed]

how can I calculate the rank score what is its syntax?

[Nick Cox]

What's your definition of a rank score? Even a glance at the help for egen shows that there are different options for ranking and beyond that you need to tell us what you understand by rank score.

[Sally Ahmed]

sorry for the repeated question I tried to delete it but I couldn't

[Sally Ahmed]

I am trying to calculate a variable called STAKE which is calculated as the mean country-level rank score of the following four indices:

Country	STAKELAW Employment Laws	STAKELAW Social Security Laws	STAKELAW Collective Relations Laws	STAKELAW Human Rights
Australia	0.35	0.78	0.37	91
Austria	0.5	0.71	0.36	95
Belguim	0.51	0.62	0.42	96

[Nick Cox]

The repeated question #4 is not a problem. Don't worry about that.

But #7 doesn't answer #5. yet.

[Sally Ahmed]

I really don't understand what the authors meant by the rank score, but they said that after getting the mean country-level rank score of the following four indices the output was STAKE and then they use the principal factor analysis to combine STAKE with other three variables to get their average.

[Nick Cox]

If you want to replicate their analysis, I think you'll need to get more details from the authors.

If you want an analysis in similar spirit, you are free to make your own decisions. If the aim is to average similar indicators, with nothing else said, then factor analysis seems to me a roundabout way to do that, but you'll find three opinions on that from any two researchers.

The range of ranks clearly depends on the number of countries (or generally observations) being ranked, which I might want to scale to

(rank MINUS 0.5) / #observations

if I were doing this myself. More at https://www.stata.com/support/faqs/s...ing-positions/