Need help; Residual Plots

[littlebecc]

Can someone give me a clear way to do them
I've completely forgotten how to do them even though i did them like last week.
My brains about to explode.
Yours sincerely, Bec

[mba]

What calculator are you using? Classpad or the TI or one?

[littlebecc]

TI 89 I think it's called..

[Gloamglozer]

TI 89 I think it's called..

Excellent, my area of expertise.

1.  Go into Stats/List Editor.
2.  In "list 1" type in your x values.
3.  In "list 2" type in your y values.

NB: In list 1, you can type in your y values and in list 2, you can type in your x values, but for simplicity, we'll just stick with steps 2 & 3.

4.  Press F4.
5.  Go down to "3: Regressions"
6.  Select "1: LinReg(a+bx)"
7.  Make sure "X List" is set on "list1" and "Y List" is set on "list2"
8.  Press enter
9.  A box should pop up informing you of some useful statistics but we want a residual plot, so press enter again.
10.  Press F2.
11.  Select "1: Plot Setup"
12.  Press F1
13.  Make sure these are the settings:

Plot Type: Scatter
Mark: Box (or anything you want but boxes are easier to see)
x: list1
y: statvars\re
Use Freq and Categories: No

NB: For the "y:" field, the easiest way is to press "2nd" and then the "minus" key, then scroll down until you see the word "resid", press enter and that should be correct.

14.  Press enter until you're back the "Plot Setup" screen.
15.  Press "F5 ZoomData".
16.  You're done.

EDIT:  Post edited for corrections.

[Martoman]

Or just so you can understand...

You are plotting the difference between the actual value and the predicted values.

So you have two variables, x and y and a table of values.

You find the equation relating x and y, say y = 4x+2

Then for your values of x, you shove them into the equation you just got.

Then you subtract these y values from the original y value.

So y actual - y predicted

This gives you really disgusting data, probs with lots of -ves and +ves.

Then you graph the original x values against these values, taking the residuals as the new y values.

If the resultant plot is random hey! you've got a good relationship there.

If it has a pattern, it suggests that you need to linearise the data with transformations.

[littlebecc]

Thanks so much Gloamglozer, you're instructions were spot on However, one question, for this bit:
5.  Select "1:Plot Setup"
6.  Press F4.

Because I press f4 and essentially 'untick' it, why do i do that?

And thanks martoman for trying to clarify, got a bit confused in the middle of your writing but i'm awful at maths (yes i know, even further) so any help i can get is awesome. Really appreciate it.
So just to clarify- a residual plot that appears random means what? That the plot is linear?! And if it has any sort of pattern you need to transform it?

God, sorry i'm so clueless when it comes to maths. Thanks again guys.

[Martoman]

A random residual plot. Like imagine you had a collection of dots in your hand and you just dropped them on the graph. It will have no pattern.

However, if a residual has a pattern you can draw in a curve at some point. Tho it can be a straight line.

http://img243.imageshack.us/gal.php?g=patternd.png

There are some pretty pictures. The ones with red lines show the general trend of the points, (that is a pattern!!!). The other is what i mean by random.

If it is random, then the equation of the line you found was a good fit!!!!

[Gloamglozer]

Thanks so much Gloamglozer, you're instructions were spot on However, one question, for this bit:
5.  Select "1:Plot Setup"
6.  Press F4.

Because I press f4 and essentially 'untick' it, why do i do that?

Sorry, my instructions were a little inaccurate at that point.  The error I made was actually in the typing - I wrote steps 12-13 twice.  If you follow the instructions now, you should be able to do it correctly.

Once again, sorry about that.  If you do follow what I had before, it just unticks it because you're telling the calculator not to plot the graph.

[littlebecc]

thanks gloam and marto............. I may use this 'asking more question things' more often You guys are good.
P.s marto, the picture didn't wory But i understand what you mean
I have 4 further SACs next week (core) and i'm kind of freaking out I missed out the class on deseasonalisation and all that stuff. Ahhhhhhh, stress.