Math 104
April 8, 2005
SPSS
Assignment 7 (due April 15)
Read most of Chapter 20 in the SPSS text: pp. 435-454.
Turn in these problems:
Chapter
20 Concepts (pages 462-465): Problem 1acd, 6.
Chapter
20 Data Analysis (pages 465-469): Problems 1abcd, 2abcde, 3ab, 5abc, 9.
HINTS
AND SUGGESTIONS
For the reading:
This is the SPSS text’s section on
regression lines. The SPSS text usually
calls them “least-squares lines,” but that’s a synonym for regression lines.
There are some other differences in
style. For the example, the SPSS text
uses “y = mx + b” for the equation of a line, while we
usually used “y = bx + a.”
That would be less confusing if the same symbol “b” weren’t being used
in two different ways. Unfortunately,
those are the two forms most commonly used.
The SPSS text talks a lot about “R
square” or R2. The “R” is the
same thing as our correlation coefficient, r.
They’re squaring it because for some purposes, they want it always to be
positive. Both forms are widely used. SPSS will calculate R2 for you,
but if you want R (or r) you have to either
(a) Take the square root of R2
by hand, and guess the sign, or
(b) Get SPSS to print out a “Model
Summary” table and look for r in it.
On page 443, there’s a formula for
the slope that looks harder than the one you’re used to. That’s because the SPSS text assumes that you
don’t know how to compute r.
Essentially, they’re doing it inside the formula. Later, on page 446, they get r from
the slope by using your formula backwards.
(Fortunately, you may never have to any either of these formulas again.)
The reading gets to be hard going,
especially around page 450. If you must
skip stuff, jump to “Some Warnings” on page 451.
For the “Concepts” questions:
Problem 1: Part (b) (the “intercept” or
constant term) is optional. Do it if you
think it helps you. You might be able to
answer the other parts without using the intercept.
If you’re doing this
in a Word document, feel free to draw the figure by hand.
Problem 6. The point is that regression
only works for numerical variables. (And
only real numerical variables—if numbers are used as meaningless labels, that
isn’t good enough.)
For the “Data Analysis”
questions:
Problem 1: Make
sure that “husbeduc” is the dependent variable, on the vertical or Y axis.
“wifeduc” belongs on the X axis.
To draw the line,
you’ll need the Chart Editor. Look for
the “Fit Line” box, and select “linear regression.” Check both the check boxes.
For problem 1c, you
can find exact values in the “Coefficients” table (see next problem). The value of R (for 1d) is in the “Model
Summary” table (also see next problem).
If you can find these tables, copy them into your word-processor
document.
Problems 2d and 2e: These
probably need to be done by hand, but you may be able to get SPSS to give you
some of the needed information. Hint on d: If the husband’s years of education were
always exactly the same as the wife’s years of education, what would be the
equation describing their relationship? Hint on e: If wife’s education were really useless for
predicting husband’s education, then what value would you always guess for
husband’s education?
Problem 3 is like problems 1 and 2, but you need to switch the X and Y
variables. That means starting over with
the regression — almost nothing is salvageable when you make the switch.
Problem 5a. “wifeft” becomes the “Set
Markers by” variable.
Problem 5b. In order to draw separate regression lines and
get separate R2 values, do the following:
Open the Chart Editor
Click on “Options” under the “Chart”
menu
Find “Fit Line” in the Options Box,
and select “Subgroups”
Click on “Fit Options..”.
Click on “Display R-squared in the
legend” and “Linear Regression”.
Click Continue and OK (but keep the
Chart Editor open for now).
Unless you’re
printing in color, you’ll have to do something to make the lines look different
from each other. In the Chart Editor,
click on one of the lines and look for the “styles” options. There’s lots of room to be artistic here. You can probably change the style of dots in
your scatterplots, too, if color isn’t enough.
Problem 9. “salnow” is the Y variable,
“salbeg” is the X variable. (That’s the
same way as in SPSS-6, but it’s more important now.)
(end)