Ch12: correlational research strategy Flashcards
What does the correlation coefficient describe?
The MAGNITUDE and DIRECTION of relationships
Pearson r is typically correlated for ___ relationships and ranges from __ to ___
Pearson r is typically correlated for LINEAR relationships and ranges from -1 to +1
What does the sign on Pearson r indicate?
DIRECTION (not magnitude) of relationships
For nonliner relationships, what coefficient would you use?
a Spearman Correlation Coefficient for non-linear monotonic relationships.
For ORDINAL DATA (eg., sex and intelligence)
When both sets of data are nominal, what statistical analysis would you use?
Chi-squared test
When one variable is nominal and the other is interval/ratio, what correlation results?
a point-biserial correlation
to establish a correlation, you need a __ of scores
range
What do you use to determine the significance of the correlation coefficient?
a t-test
What is R^2?
- coefficient of determination
- proportion of variability in the DV accounted for by the IV
For rest-retest reliability, what ccorrelation would you use?
Pearson r
For split-half reliability, what correlation would you use?
Spearman-Brown formula
-note: first calculate R, then Spearman-Brown formula
The ___ the correlation, the better the split-test reliability
higher
What are correlations used for in inter-rater reliability?
Cohen’s Kappa is used
Cohen’s kappa = (%agreement - percent chance) / (1-%chance)
What is Cohen’s Kappa?
measure of agreement that attempts to correct for chance
Without some correction, measures of inter-rater reliability…
Tend to overestimate the true level of agreement between the raters
What does convergent validity seek to answer?
are 2 measures of the same construct correlated?
___ validity is seeking to make sure that 2 measures of different constructs are NOT correlated
divergent
____ is used to test significance of r value
t-test
cyclical functions make it difficult to determine ___ ___
temporal precedence
We use ____ to predict scores on one variable based on scores on another variable
REGRESSION
In a regression, what is equivalent to the DV and the IV?
DV = criterion variable IV = Predictor variable
Degrees of freedom for t-test on significance of Pearson r value
N-2
With 2 variables, what regression do you use?
simple linear regression
Regression equations
y = b1x + b0
- where y has a cap (^)
- b1 is the slope
- b0 is the intercept
- like y=mx+b
What is the role of the line of best fit in a simple linear regression?
to minimize the distance of each data point from the predicted value
T/F: the size of b1 in the simple linear regression equation (slope) tells us about the strength of the relationship
FALSE. b1 tells us the scale, but doesn’t say anything about the strength of the relationship
-as x increases, by how much does y change?
Purpose of a regression equation?
To make predictions based on the line of best fit
What does the sign of b1 tell us?
the direction of the relationship
Purpose of a multiple regression
to maximize predictive power.
-more variables = more predictive power
Multiple regression equation
y = b1x1 + b2x2 +b0
How could you maximize predictability in the following example:
Likeability is predicted by physical attractiveness of politician
- add in a measure of political affiliation
- more variables = higher predictive power
How do you test the significance of regression equations?
by using the standardized beta.
- Standardized beta tell you the SIZE OF THE RELATIONSHIP. It tells you the change in SD units.
eg. , as x variable changes by 1 SD, by how many SD’s does the y variable change? By insert standardized beta value
What does the correlation of determination (R square) tell you about a regression equation?
The proportion of variability explained by the regression line
In an SPSS output, 4 variables to predict murder rate are in a table.
The standardized beta for resident population is 0.052 and for burglary rate per 100,000, it’s 0.333. For lacency-theft rate per 100,000, its -0.205 and for motor vehicle theft rate per 100,000, it’s 0.0507. Explain what this means
- As population increases by 1 SD, murder rate goes up by 0.052 of a SD. This is a VERY WEAK relationship.
- A stronger relationship exists with burglary rate (0.333)
- as larceny increases by 1 SD, murder rate decreases by 0.205
- as motor theft rate increases by 1 SD, murder rate increases by 0.507 SD. This is our STRONGEST PREDICTOR. Its significance (p value) is <0.001.
If the significance for resident population is 0.576, what does this tell us?
that resident population is not a significant predictor of murder rates
Correlation, simple linear regression or multiple regression? For:
A local gelati store tracks the number of bowls they sell and the daily high temperature each day for several weeks so they can budget and stock for next summer.
Simple linear regression, because there are 2 variables (one PREDICTING and one criterion)
Correlation, simple linear regression or multiple regression? For:
Researchers are interested in determining the relationship between death anxiety and religiosity
Correlation
Correlation, simple linear regression or multiple regression? For:
A university wants to know how accurately scores on each sub-component of the GRE can determine the success of a graduate student.
Multiple regression. PREDICTING future success based on multiple subcomponents of the GRE
What kind of reliability analyses can you run with correlations?
- test-retest reliability
- split-half reliability
- inter-rater reliability
What kind of validity analyses can you run with correlations?
- Convergent
- Divergent validity
