W12 non-parametric tests

Question 1

if assumptions violated: not normally distributed…

Answer 1

use non-parametric

Question 2

Null vs alternate hypothesis for CORRELATIONS

Answer 2

pearsons r null = no relationship
alternate =there is relationship

Question 3

non-parametric pros + cons

Answer 3

pros = robust, ease of interpretation
cons = less power, limited precision

Question 4

NON-PARAMETRIC tests include:

Answer 4

Chi-square = for t-test, ANOVA
Spearman’s Rho = correlation

Question 5

z,t,ANOVA rely on PARAMETRIC assumptions like:

Answer 5

1. continuous - interval or ratio
2. normally distributed
3. homogeneity (levenes) or sphericity (muachelys)

Question 6

Assumptions are violated if…

Answer 6

1. non-normally dist (skewed, bimodal)
2. categorical (nominal or ordinal)
3. variances wildly dissimilar across IV

Question 7

Violating assumptions = bad because…

Answer 7

1. not real result – false positive type 1 error
2. miss a result – false negative type 2 error
3. unreliable pvalue/intervals

Question 8

Chi-Square Xsquared

Answer 8

tests if theres a significant difference in frequency we’d expect, and observe to happen

Question 9

Two types of Chi Square

Answer 9

1. Goodness of fit - if single cate variable fits in expected distribution
2. test of independence - if 2 categorical variables are related

Question 10

Chi Square
goodness of fit H0 and H1

Answer 10

H0 null is observed = expected
H1 is observed is NOT = expected

Question 11

GOOD FIT formula

find value of xsquared then go to table

df = k(number of groups) -1

Answer 11

df on side column, 0.05on row
if obtained is bigger, reject null hypothesis

Question 12

INDEPENDENCE

when you know observed but not expected – need to calculate it

Answer 12

row total x column total
divided by
grand total all cells

Question 13

Assumptions for Chi Square:

Answer 13

• categorical data
• independence of observation - each participant appear once
• expected frequencies (at least 5 each cell)
• sample size - large

Question 14

Spearmans’ Rho

Answer 14

• nonparametric
• measure strength, direction of association between 2 variables
• data = ranked Ordinal likert scales
• range -1 to 1

Question 15

RULES: When to use spearman’s rho

Answer 15

1. at least one variable is ordinal
2. data not normally distributed or has outliers
3. relationship is non-linear

Question 16

non-linear monotonic relationship is

Answer 16

• moves in one direction (not at constant rate)