8. Statistical Validity (Pearson Assumptions)

Construct validity

How well are variables measured

Interval validity

Was the goal to find a causal relationship

External validity

Is the sample random, can we generalize to pop.

Statistical validity

Combines significance (p-val), suitability, relevance, accuracy

Assumptions Pearson’s r

Sample is random, both variables are interval/ratio, rel. is linear

If values are ordinal

Use Spearman’s correlation (1)

If values are nominal

Use Chi-square test of interdepence

Chi-square test of interdependence

Test that det. the degree to which distribution of 2 nominal variables is dependent on one another

If Chi-square variables are independent

Rel/ proportion would hold regardless of categ (have similar results)

If rel. is monotone curvilinear

Use Spearman correlation (2)

One-sided hypothesis

Assumes direction of correlation (H_A: rho>0/ rho<0)

Adv & disadv of one-sided hypothesis

(+) theory-driven; (-) if rel turns out to be opposite dir, you can’t reject H₀

Two-sided hypothesis

Says there is a rel. bet. var, but does not specify direction (H_A: rho =/= 0)

Adv & disadv of two-sided hypothesis

(+) looks at poz and neg possibilities; (-) does not match expectation, less likely to adopt new theory