Class Notes
(809,509)

Canada
(493,754)

University of Saskatchewan
(2,893)

Sociology
(204)

SOC 325
(8)

Lecture

# March 24(4).docx

Unlock Document

University of Saskatchewan

Sociology

SOC 325

Elizabeth Quinlan

Winter

Description

Soc 325
March 24 2014
1
-- Lab next Monday in Arts 40, basement computer lab!
-----------------------------------------------------------
• Correlation: (Pearson R) measure of the strength of a relationship between 2 interval-
ratio variables
• Regression: a technique that gives us the form of the relationship between 2 interval-
ratio variables, X and Y, that bests predicts values of Y based on values of X
o Regression is not symmetrical, X and Y, so if you reverse the independent and
the dependent you will get a different picture
o Generally a rule for sorting the form, which comes first in time?
First: independent
Second: dependent
Assumptions of correlation and regression
1. Both variables are normally distributed
2. The relationship is linear and homoscadastic
o Homoscadastic: general shape forms a cigar
o The further from the shape the more tentative we have to be in the language
about conclusions
"we see the following with caution ….. "
• Creating scatterplots in SPSS
o https://www.youtube.com/watch?v=H7Fz1dCiLZk
SPSS
• Correlation:
o Analyze -> correlate --> choose bivariate --> put in variables --> click Pearson r
(which is the default) --> produces a correlation matrix (upper right mirrored in bottom
left) --> Pearson's r .389 so there is a strong relationship (larger than .3) --> r-squared =
0.151
Job tenure and usual hourly wages -- variables
• Regression:
o Analyze --> regression --> linear --> enter independent and depend separately
(hourly wages = dependent, job tenure = independent) --> constant = a
Y = a + bx
Y = constant + (second line)x
Y = 18.22 + .051X
A = rate of pay for someone with no job tenure
Average starting wage
Value of y when x=0
B = increase in dependent value for every unit of increase in the
independent variable
5 cents for every additional month for being on the job
We can use the regression equation to predict the starting wage
for someone at any number of months
Example, someone who worked for 30 months
Y = 18.22 + 0.051(30) = 19.75 Soc 325
March 24 2014
2
Don't have to have a person in the data set to know what their
wage will be/hour
o More summary: shows are r and r-squared --> we can use for a PRE measure
The adjusted R-squared sometimes differs when we work with multiple
regression (more than one independent variable)
o Std: error of estimate: the std deviation when the independent variable is moved
o Anova: reports f-statistics (test statistic for regression) to determine if coefficients
are significant or not
Significance: less than 0.05
Means that the coefficients of the regression model are statistically
significantly different from zero
f-statistic is the ratio of the 2 numbers from the column before (mean-
squared top/mean-square bottom)
Sum of squares total = error one
Sum of squares f residual: error two
Error one - error two = regression
r-squared = explained/total
Chapter 14:
• Three and more variables
o Multivariate analysis
1. Partial correlation
Types of outcomes we can expect
Will the relationship between X and Y retain its strength
and direction after introducing another variable, Z?

More
Less
Related notes for SOC 325