y predicted values given x

[ast972]

Good morning all,
I just installed this interesting package, and I find the PPS interesting and well motivated for feature selection. Given a vector y, and a vector x, you tell us whether x has 'explanatory power' relative to y, measured between 0,1. And viceversa. So given any attempt to write a causal model for y, we should somehow consider x.

You are fitting a model behind the scenes, that you choose from a menu of models (and if I understand correctly you focus on the decision tree class). But then, I wonder how can I see the 'predicted' values for y (what econometricians call yhat) given the variables x (what econometricians call 'explanatory variable') for the given sample (y,x). Next would of course be the model 'yhat' for a fresh x, not in the sample.

Perhaps this is buried in the dict output but I couldnt find it. Could you help?

Thank you

Angelo

[8080labs]

Currently, yhat is not returned. What is the background of your request? In which situation do you want that? Sent from mobile

…

On 7 Jan 2023, at 12:40, 'ast972' via Admin ***@***.***> wrote:  Good morning all, I just installed this interesting package, and I find the PPS interesting and well motivated for feature selection. Given a vector y, and a vector x, you tell us whether x has 'explanatory power' relative to y, measured between 0,1. And viceversa. So given any attempt to write a causal model for y, we should somehow consider x. You are fitting a model behind the scenes, that you choose from a menu of models (and if I understand correctly you focus on the decision tree class). But then, I wonder how can I see the 'predicted' values for y (what econometricians call yhat) given the variables x (what econometricians call 'explanatory variable') for the given sample (y,x). Next would of course be the model 'yhat' for a fresh x, not in the sample. Perhaps this is buried in the dict output but I couldnt find it. Could you help? Thank you Angelo — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.

[ast972]

indeed, when we estimate a correlation rho between two normalized random variables (x,y), we can approximate the conditional expectation of y given x as yhat = rho*x. this is useful in many applications, the most obvious is linear regression. Is there an equivalent in ppscore measure?

[8080labs]

Understood. There is no equivalent available. If you want to get a prediction for a certain value, you might want to train your own decision tree(s).Sent from mobileOn 24 Jan 2023, at 21:37, 'ast972' via Admin ***@***.***> wrote: indeed, when we estimate a correlation rho between two normalized random variables (x,y), we can approximate the conditional expectation of y given x as yhat = rho*x. this is useful in many applications, the most obvious is linear regression. Is there an equivalent in ppscore measure? —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: ***@***.***>