Linear Regression - Least Squared Distances

I see what your saying, given that you remove outliers, the resulting 2 lines should be pretty similar too.

if you remove the circled outliers the lines are quite close as you suggest.

http://i54.tinypic.com/ilwutd.jpg

The risidual distribution (right hand graph) isn't really as predicted but this appears to be due to the heavy rounding of residuals before they are plotted.

Quote:

---

Originally Posted by Corpsecreate

Isn't "Best Fit" used to describe the line that will come "closest" to intersecting the data points?

---

Yes, given a measure of fit. Without a given measure of fit, there is no useful meaning.