Correlations Part 2

Question 1

Significance tests

Answer 1

Parametric (pearson’s product moment) vs non parametric (spearman’s rho).
P- how likely relationship is due to chance.
Alpha (a)- level we decide relationship isn’t due to chance.
Significant- p.05; fail to reject Ho.
Never proves a relationship.

Question 2

Parametric assumptions

Answer 2

Data should form normal distribution.
Kolmogorov-smirnov result typically comes from interval data.
Assumption of independence- behaviour between ppts should be unrelated.

Question 3

Spearman’s rho (non parametric)

Answer 3

Use when:
Ordinal data- X and Y are ranked.
Interval data that doesn’t meet parametric assumptions.

Logic:
Rank sets of numbers. 
Identical- create rho of +1. 
Opposite- create rho of -1. 
If there are ties- share ranking value.

Question 4

Pearson’s product moment (parametric)

Answer 4

Variance formula:
S^2X = sum(X-meanX)^2/N-1.
CovXY = sum(X-meanX)(Y-meanY)/N-1.
r = covXY/SxSy.

Logic:
CovXY- deviations from the mean; multiply to create covariance.
Divided by individual variance- r.

Question 5

Hypothesis testing

Answer 5

Null hypothesis- any relationship is due to chance.

Experimental hypothesis- non directional/two tail; directional/one tail.

Question 6

Comparing spearman’s and pearson’s

Answer 6

Same data, two tests.

Parametric statistics are more powerful because of restrictions to the data.

Question 7

Relation to sample size

Answer 7

Correlation measures the degree two sets of related scores follow the same pattern.
Could be by chance; more likely with less ppts.
Testing significance requires knowing sample size and coefficient.

Question 8

More types of correlations

Answer 8

Kendall’s tau- non parametric; small data set with lots of tied ranks.
Point-biserial- when one variable is dichotomous (two distinct parts eg. male or female).
Biserial- one variable is continuous dichotomous.