In an A-Level maths textbook that I am reading about least-squares regression, it states that a regression line should only be used to predict the dependent variable and not the independent variable. I am looking for some intuition about this.
Naively, I would expect the regression line of $y$ on $x$ to be an inverse of the regression line of $x$ on $y$ . Specifically, I would expect the model $y_i = a + b x_i$ to produce an inverse model $x_i = -\frac{a}{b} + \frac{1}{b} y_i$.
Why is this not the case?