Correlation and Regression Lab Guide
Uploaded by
PohuyistCorrelation and Regression Lab Guide
Uploaded by
PohuyistName: _______________________________________________________
PSYCH248 Computer Lab Lab Week 12 Correlation & Regression
PART 1: BASIC CONCEPTS in CORRELATION and REGRESSION.
A. Please go to website: [Link]
Click on “Show Movable Line”
Manipulate the line through the data that you think best fits the entire set of data.
Then click on the “show regression line” and “show residuals” on the right side.
Compare your Sum of Absolute Errors (SAE) on the right side with the SAE derived from the
actual line of best fit. The SAE is the sum of absolute errors between the predicted and the
eventual-real outcomes.
Compare your line to the actual regression line. How far off were you in drawing the line of best?
Subtract your SAE from the SAE on the right:
___________________________________________
Lower SAE scores mean what?____________________________________________________
What do you think the correlation coefficient will be? _______ What is the actual score?______
Unclick the “Show movable line” button and click the “Move Observations” button.
The Residual score is the variation in Y not related to changes in x.
Does this represent the null or the alternative hypothesis?
____________________________________
What do you have to do to the data points to make the residual score come down to zero?
_________________________________________________________________________
______________________________________________________________________________
What will this do to the correlation score? ___________________
What is the reason for both of these changes?
______________________________________________________________________________
______________________________________________________________________________
What name do we give this variability? ___________________________________________
PART 2: Using SPSS to generate a correlation matrix and regression output.
1. Open [Link] on the server K drive. Use Analyze, Correlate, Bivariate to create a
correlation matrix with these 5 variables: height, weight, waist circum., pulse rate, BMI
How many unique correlations were calculated? ____ (note that the matrix contains duplication)
What is the Pearson correlation between: height & weight? _____ height & waist size? _____
weight & waist size? _____ pulse rate & BMI? ______ pulse rate & height? _____
Mark the significant correlations above with a * if sig <.05, ** if sig <.01
2. Which correlation is negative? ___________________ Describe what that relationship means.
Of all of the correlations above, which is the strongest?_______ which is the weakest?_______
Summarize the patterns found in the significant correlations - in simple terms.
______________________________________________________________________________
______________________________________________________________________________
3. Run a Regression equation. (Use Analyze, Regression, Linear, specify weight as the
dependent variable and height as the independent.) Look first at the Coefficients table:
Find the slope (B): ______. This means as height increases 1 inch, weight increases 4.5 lbs.
The Constant listed above the slope for height is the Y-intercept. This is the height predicted (Y) if
weight (X) were 0. We get a negative value --so we can’t extrapolate down to zero.
b. Look at the Model Summary table: (see R Square value).
What % of variance in weight does height account for? ____________ or ___________%
c. How far off on average would your estimate of weight be if you used only height to predict it?
________________ (see standard error of the estimate)
d. Here is the regression equation for this analysis. Weight = 4.553*height + (-138.6)
Use this to make predictions:
For a man who is 66” what is his predicted weight? _________
For a man who is 72” what is his predicted weight? _________
4. Run another Regression equation with the same DV but using waist circ. as the independent.
a. Find the slope (B): ________. This means as waist increases 1 cm, weight increases ____ lbs.
b, What proportion of variance in weight does waist size account for? _________ or _______%
c. How accurate would your weight estimate be if you used only waist to predict it?__________
d. Write the regression equation for this analysis. Weight = _______* Waist + (___________)
Use this to make a prediction: A man with waist = 100cm has predicted weight = _______lbs?