1

I came across this equation when I was reading the article: $$\textrm{Y} = \alpha + \textrm{X}_1 (\beta_0 + \beta_1 \textrm{X}_2 + \beta_2 \textrm{X}_3) + \varepsilon$$

At least, I can understand the latter two variables in the parenthesis: both $\textrm{X}_2$ and $\textrm{X}_3$ affect $\textrm{Y}$ with $\textrm{X}_1$. However, there are several things do not make sense to me:

  1. How can I interpret $\textrm{X}_1 * \beta_0$? Is it fine to say constant affects $\textrm{Y}$ with $\textrm{X}_1$?
  2. When we use interaction term, don't we have to include the variables with no interaction? Shouldn't be the equation like this? $$\textrm{Y} = \alpha + \textrm{X}_1 (\beta_0 + \beta_1 \textrm{X}_2 + \beta_2 \textrm{X}_3) + \beta_3 + \beta_4 \textrm{X}_2 + \beta_5 \textrm{X}_3 + \varepsilon$$
user51966
  • 245
  • 2
  • 9
  • 1
    Is $X_1$ a binary variable or something else? – whuber Jun 17 '15 at 22:40
  • 1
    It is an interval variable. In the article, it is defined as "effective number of elective parties". – user51966 Jun 18 '15 at 00:28
  • 2
    Set $X_1$ to zero & the model reduces to $Y=\alpha+\varepsilon$: there is no relationship between $Y$ & $X_2$, or between $Y$ & *$X_3$ for this special value of $X_1$. The general issues with such constraints are discussed [here](http://stats.stackexchange.com/q/133959/17230), [here](http://stats.stackexchange.com/q/27724/17230), & [here](http://stats.stackexchange.com/q/11009/17230); you'll have to include some context to allow anyone to say whether it makes sense in this particular case. (BTW the alternative you suggest appears to have an unnecessary $\beta_3$ on its own.) – Scortchi - Reinstate Monica Jun 18 '15 at 09:48
  • 1
    It may help to understand the meaning and purpose of the model. Since $X_1$ is some kind of count or proportion, the model seems to propose that $Y$ is the product of two quantities (plus an additive offset and error), which might make perfectly good sense as a count weighted by something else or as a proportion of something. – whuber Jun 18 '15 at 14:18

0 Answers0