Estimate nonlinear correlations between two variables

[patrick handcock]

Hi folks,

Apologies for beginner question.

In R, it is possible to estimate nonlinear correlations between two variables (e.g. https://towardsdatascience.com/estim...r-62c6571cb1db).

Is there a similar function in Stata? I was a little unsure, but my thinking was:

Perhaps one way could be to use regression with some splines and extract R^2 (the squared correlation), and taking the square root (though that loses the sign, as I think does the approach in nlcor). The sign would be interpreted from the coefficients / plots? Does anyone have potential code to do this, or perhaps there is a simpler way?

Many thanks.
Patrick

[Carlo Lazzaro]

Patrick:
you may want to take a look at -spearman- and -ktau-.

[patrick handcock]

Carlo Lazzaro - many thanks. I may have misunderstood, but are spearman and ktau not largely limited to linear estimates (i.e. using rank order involves a loss of information about distance between observations, so cannot robustly estimate nonlinear correlations)?

[Carlo Lazzaro]

Patrick:
as second thought, see Richard Williams 's excellent teaching notes on Nonlinear relationships at https://www3.nd.edu/~rwilliam/stats2/l61.pdf.

[Nick Cox]

See also FAQ https://www.stata.com/support/faqs/s...ics/r-squared/