What’s the Difference Between Y-on-X and X-on-Y Regression?

The regression line of $y$ on $x$ predicts $y$ from a known value of $x$, whereas the regression line of $x$ on $y$ predicts $x$ from a known value of $y$. They are generally different equations and are not obtained by simply rearranging one another.

Why the Lines Are Different

A regression line uses the least-squares method to minimize prediction errors, called residuals. However, each line minimizes errors in a different variable.

The variable placed after “on” is the known explanatory variable, while the variable before “on” is the response being predicted. Thus “$y$ on $x$” means regress $y$ using $x$, not place $y$ on the horizontal axis merely because of graphing convention. Always read the phrase as “predict the first variable from the second.”

For example, suppose technology gives:

$$y=2.1x+4.3$$

This is the regression line of $y$ on $x$. If $x=10$, then

$$y=2.1(10)+4.3=25.3.$$

Suppose the regression line of $x$ on $y$ is

$$x=0.42y-1.5.$$

If $y=20$, it predicts

$$x=0.42(20)-1.5=6.9.$$

A common misconception is that rearranging $y=2.1x+4.3$ produces the regression line of $x$ on $y$. Rearrangement gives the same geometric line, but the actual $x$-on-$y$ regression is calculated separately because it minimizes different residuals. The two regression lines coincide only when there is perfect linear correlation, .

Both fitted lines pass through the mean point $(\bar{x},\bar{y})$, but unless $|r|=1$, their gradients reflect different optimization problems. A strong correlation can make them look similar, yet that does not make one the algebraic inverse of the other. Use the appropriate GDC output.

IB Exam Technique

This is Mathematics AA SL 4.10 and is normally assessed on Paper 2 using a GDC. Identify the variable being predicted before selecting the equation: predict $y$ from $x$ with the $y$-on-$x$ line, and predict $x$ from $y$ with the -on- line. Also avoid unreliable extrapolation beyond the observed data range.