(How) Can I calculate a significance test for Kendall´s Tau c in Stata?

[Laura Bartz]

Dear community,

I calculated Kendall´s tau c in Stata "by hand" because the ktau-command only contains results for tau a and b.
I used the syntax for tau c provided by Roger Newson (https://www.stata.com/statalist/arch.../msg00300.html).
Now I am wondering if there is a way to calculate a significance test or even better, a confidence intervall, for tau c.

Thank you in advance!
Laura

[Nick Cox]

Bootstrap it?

[Laura Bartz]

Yes, I tried that but I didn´t find the right syntax yet. I am not that experienced in the use of Stata. Do you maybe have a clue for me how to use the bootstrap-command in this case?

[Nick Cox]

You'd need to write a wrapper program that emits tau_c as a r-class or e-class result. Sorry, but that is a bigger deal than I have time for today. Someone else may be able to help.

[Laura Bartz]

I see, thanks for your hint. I will try to do something with it myself and hope for someone being able to help.

[FernandoRios]

Hi Laura
The Bootstrap procedure that i use for cases something like this is:

program myktau, eclass
syntax varlist
ktau `varlist'
matrix b=[r(tau_a),r(tau_b)]
** This is just for illustration, and should follow whichever formula you have been applying
local tau_c=r(tau_a)*r(tau_b)
matrix b=b,`tau_c'
matrix colname b=tau_a tau_b tau_c
ereturn post b
end
sysuse auto, clear
bootstrap:myktau foreign trunk

This should do what you need to obtain CI, as long as you replace tau_c with the correct formula, which i think you already have.
This is also a very simple version of the bootstrap that uses all data. You may want to adjust it for specific cases.
HTH
Fernando

[Laura Bartz]

Dear Fernando,
yes, I do have a formula for tau c that calculates the value of it and I have specified datasets which only contain the cases I want to include in the bootstrap. So I think your suggestion will be very helpful once I manage to apply it to my data. Thank you very much!
Cheers
Laura