0
$\begingroup$

I have no idea about how to tackle a problem with 2 quantitative variable X and Y. How can i model $Y=f(X)$ ? What is $f$ if i just have a look of the data ? how can i even guess if such an $f$ exist ?

I feel like i have no key to determine a relationship and the kind of relationship between two variable... (and i don't find anything in Hasti Tibshirani 's book)

For example i try to manage a complete data project. One question asked is to "determine the dependance between two variable: "age" and "note" (grade)".

I have 2 quantitative variable so:

  1. I first plot them (the colours distinguish the class variable: 0 and 1 but is not important here!) We can see that there isn't any linear correlation on the graph. Moreover i can't see any other kind of relationship (like log aor squared root, polynomial,...)

  2. I didn't know what to do to "evaluate" the "dependance" between my two variable so i did a linear model and without any surprises the coefficient $R^2$ is really small wich indicate a really poor linear relationship.

I don't know what to do next ...

If you have any suggestions please share it.

Thanks

$\endgroup$
  • 1
    $\begingroup$ Chapter 3. Linear Methods for Regression (The Elements of Statistical Learning) $\endgroup$ – AidanGawronski Apr 12 '18 at 16:14
  • $\begingroup$ I am sorry but i have the book and can't find any suggestion in the chapter 3 about choosing $f$ ... $\endgroup$ – laurent pincemaille Apr 12 '18 at 17:35
  • $\begingroup$ You don't choose the function, you fit the regression line, and the coefficients which minimize the sum of squared errors are the function. $\endgroup$ – AidanGawronski Apr 12 '18 at 18:00
  • $\begingroup$ Yes i fear that i was not clear enough, my main concern is to determine a candidate for $f$, and i wanted to know if an automatic procedure could achieve it? $\endgroup$ – laurent pincemaille Apr 12 '18 at 18:39
  • $\begingroup$ Imagine we have a big dataset (lots of variable) and we have to predict a quantitative variable Y. Hence we want to know wich independant variable Xi are relevant. If we have more than 2 independant variable we can' t make a graph and guess intuitively from the graph wich f could convey Y=f(X1,...,Xn) . $\endgroup$ – laurent pincemaille Apr 12 '18 at 19:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.