lecture 14 Flashcards
(21 cards)
correlational research
- describe and or predict behavior
ex) relationships between variables - correlation does not imply causation
measuring the strength of relationship
- correlation coefficient describes the magnitude and direction of relationship
- Pearson r is typically calculated for linear relationships
- ranges from -1 to +1
- sign (+/-) indicates direction, not magnitude of relationships
Spearman brown correlation coefficient
- Spearman Brown correlation coefficient
- non-linear monotonic relationships-less sensitive to linearity than Pearson
- ordinal data (ranked)
Point-biserial correlation
- Point-biserial correlation
- one variable is nominal (dichotomous), one variable is interval/ratio
ex) sex and intelligence
Chi-square
- Chi-square
- both variables are nominal
what do you need to establish a correlation
-a range of scores
testing significance
- you use a t-test to determine the significance of the correlation coefficient
- coefficient of determination R2
- proportion of variability in one variable predicted by other variables
what else are correlations used for
- Reliability analyses
2. Validity analysis
reliability analyses
- Reliability analyses
-measurement is consistent
a) test-retest reliability
-Pearson r
b) Split-half reliability
-Spearman-Brown Formula
c) inter-rater reliability
-Cohen’s Kappa (498-490)
how do we calculate PA and PC?
-proportion of times they both agree
-likelihood that they agree just by chance
validity analyses
- Validity analyses
a) Convergent validity
- two measures of the same construct are correlated
b) Divergent validity
- measures of different constructs are not correlated
why does correlation not equal causation
- alternative explanations
- temporal precedence
calculating a correlation
- Calculate the deviation of each score from it’s mean
- Calculate the cross-product of the deviation
- Calculate the squared values of the deviation
- Plug calculations into the formula
testing significance of the correlations
-t-test to test the significance of the r value
predicting behavior
- we use regression to predict scores on one variable based on scores on another variable
1. criterion variable= dependant variable
2. predictor variable= independent variable
ex) time spent studying and student grades - with two variables =simple linear regression
simple linear regression
-regression equations
-algebraic description of relationship between variables
-line of best fit
y=mx+b
y(hat)=b1x+b0
b1= slope
b0= intercept
line of best fit does ?
-line of best fit minimizes distance of each data point from the predicted value
what is b1
- slope of the line
- as X increases, by how much does Y change?
- what does the SIZE of b1 tell us? Nothing
- what does the sign of b1 tell us? If it’s pos or negative slope
example using equation
y(hat) =1.875x + 1.5
y(hat) =1.875(4) +1.5
y (hat) = 9
multiple regression
- multiple predictor variables
- similar to simple linear regression but now we have multiple variables
- equation that maximizes predicative power
ex) likeability predicted by physical attractiveness of politician - add in measure of political affiliation
- including more variables would mean: more predictive power
testing significance of regression equations
Testing significance of regression equations
ex) predicting exam score from hours studied
- coefficient of determination
- proportion of variability explained by the regression line
- use an ANOVA to test
standardized beta
-the rate of change, as x changes by 1 SD by how many SD does the other variable change
