8

My question is about the general difference between the interaction terms $x_1x_2$ and $x_1/x_2$ in multiple linear regression.

Suppose you are performing multiple linear regression and you have interactions present i.e. an independent variable has a different effect on the outcome depending on the values of another independent variable. Therefore you decide to model as follows:

$$ Y = \beta_0 + \beta_1x_1 + \beta_2 x_2 + \beta_3x_1x_2 $$

For example where $Y$could represent Impurity after a chemical reaction, $x_1$represents Reaction Temperature, $x_2$ represents Reaction Time.

Adding an interaction term between Reaction Temperature and Reaction Time $x_1x_2$ would effectively change the gradient for Reaction Time as we change the value of Reaction Temperature from low to high (and vice-versa for Reaction Temperature).

My questions are:

  1. What impact does the interaction term $x_1/x_2$ have when compared to $x_1x_2$ ?
  2. In what circumstances might you include an interaction term of $x_1/x_2$ as opposed to $x_1x_2$ ?
Karolis Koncevičius
  • 4,282
  • 7
  • 30
  • 47
Mike Tauber
  • 807
  • 1
  • 5
  • 14
  • 1
    Here are some related posts: https://stats.stackexchange.com/questions/58664/ratios-in-regression-aka-questions-on-kronmal, https://stats.stackexchange.com/questions/129981/interactions-using-ratio-of-variables, https://stats.stackexchange.com/questions/112878/ratio-of-explanatory-variables-in-multiple-regression, https://stats.stackexchange.com/questions/47222/techniques-for-analyzing-ratios – kjetil b halvorsen Mar 02 '21 at 12:52

0 Answers0