# Residual plots.

• Jun 27th 2009, 02:46 AM
luysion
Residual plots.
hey guys, just working on residuals. first post and hopefully not the last :P
so here the question much appriciation to however so helps :)
note- im finding it extremely hard to intrepret which is of random scatter and which is of a pattern.

which of the following residual plots shows that a linear model (the model used in the first place) is inappropriate to use to display the correlation between the two variables.

or

by the way hello http://mathguru.freeforums.org/image...on_e_smile.gif gd forums
• Jun 27th 2009, 07:27 AM
apcalculus
Quote:

Originally Posted by luysion
hey guys, just working on residuals. first post and hopefully not the last :P
so here the question much appriciation to however so helps :)
note- im finding it extremely hard to intrepret which is of random scatter and which is of a pattern.

which of the following residual plots shows that a linear model (the model used in the first place) is inappropriate to use to display the correlation between the two variables.

or

by the way hello http://mathguru.freeforums.org/image...on_e_smile.gif gd forums

A residual is a vertical difference between the observed and predicted y-values. When the residual plot is spread somewhat uniformly above and below the horizontal axis (y=0) that's an indication that the regression line fits the "cloud" of data well.

In your choices, which you would say matches the criteria better?

I hope this help!
• Jun 27th 2009, 03:17 PM
luysion
hey thanks for the reply, i want to say B.. but they both look sort of like they have a decreasing gradient?
• Jun 27th 2009, 06:13 PM
apcalculus
I would have picked the second one as the better fit since for most of the observations the residuals (between, say x=1 and 3) lie as much as above as they do below the horizontal axis. If you can, visualize a horizontal band or strip above and below the horizontal axis.

The first scatterplot seems to have a lot of positives and a lot of negatives isolated in different parts, which suggests that the line doesn't quite pass through the cloud of data, but it divides it somewhere in the middle, so the overall pattern is not captured or described by the regression line.

Good luck!!
• Jun 27th 2009, 07:33 PM
luysion
hey sorryapcalculus
i actually have an exam on this tommrow.

so i was actually thinking theoretically both graphs show an overall random scatter with no vivid pattern so to be honest dont both sort of suggest that a linear pattern may be appropriate?
• Jun 27th 2009, 10:47 PM
CaptainBlack
Quote:

Originally Posted by luysion
hey guys, just working on residuals. first post and hopefully not the last :P
so here the question much appriciation to however so helps :)
note- im finding it extremely hard to intrepret which is of random scatter and which is of a pattern.

which of the following residual plots shows that a linear model (the model used in the first place) is inappropriate to use to display the correlation between the two variables.