## Two-variable linear regressions

In the first assignment, we asked you to use several SPSS commands to describe, compare, and begin to
make inferences about two feeling thermometer variables in your ANES2016 data set. In the second, you
used T‐TESTstoconduct inferentialstatistical hypothesistests of differences on those variables. In the third,
you continued your analysis with a CROSSTABS procedure.
In this assignment, you will learn how to analyze your hypothesis using the full value of interval measures
and their associated measures of association–REGRESSION.
To make sure that you cover all that you will need to accomplish in this assignment, I suggest checking off
each question or request as you complete it.
• ‘thoroughly discuss,’ ‘discuss in detail’ or ‘fully compare’ means exactly what each implies.
NOTE: The latest versions of SPSS are a bit more finicky about the sequence of the REGRESSION syntax
command (different from p. 97 order). The sequence must be as follows. Of course, if you are using the
pull-down graphic menu, this will not matter. This is for Task 4.
REGRESSION VARIABLES=………../
SELECT=………/
DEPENDENT=……../
METHOD=ENTER.
NOTE: correction on bottom of SPSS Manual, p. 95. ‘44.911%’ should be ‘44.711%’
• Review the 3 Regression Videos—these are mainly a repeat of lectures
o Regression-1 http://www.screencast.com/t/pwZYlnbo6P
o Regression-2 http://www.screencast.com/t/1jxTsyJUx9S
o Regression-3 http://www.screencast.com/t/iKOrbVVU
• Open your ANES2016 revised data file
• As before, WEIGHT by PW2016_FULL
New Procedure – REGRESSION — SPSS Manual section 4.6
• As in HW 1 and 2, use variables V16 and V17, not the recoded v16A and v17A you created in HW 3.
• Perform 2 two-variable linear regressions
 Interpret the intercept, slope and R-square values
 Interpret the intercept, slope and R-square values
o Which is the better predictor—V18 (for Clinton) or V19 (for Trump)
 Why? What values are you comparing?
 Suggest why one is a better predictor than the other
Same Procedure, but multiple regression and use of a ‘dummy’ variable.
• Just as with an ordinal interpretation, a dichotomy can be used in an interval analysis (see
explanation in Text, p. 181)
• Perform two 3-variable linear regressions
 Interpret the intercept, slopes and R-square values
 How much does your chosen IV add to explaining the variance of V16 (above V18)?
o Which IV explains more of the variance of the DV?
 Why? What values are you comparing?
 Interpret the intercept, slopes and R-square values
o Which IV explains more of the variance of the DV?
 Why? What values are you comparing?
 How much does your chosen IV add to explaining the variance of V17 (above V19)?
Same Procedure, but 2-variable regression and your IV investigated for specification effects. Just as with
Simpson’s Paradox, relationships of subsets can be hidden.
• Perform 2 two-variable linear regressions with control
but separately for each category of your IV (Manual, pp. 97-98—see change above)
 Interpret the intercept, slope and R-square values for each of your IV categories
 Compare the two—for which of your IV categories is V18 a better predictor of
V16?