Quiz #3

Question 1

ANCOVA what it does

Variables used

Answer 1

shows the effect of the independent variable after covariates have been removed or accounted for

Uses 2+ categorical independent variable and 1 continuous dependent

Question 2

ANCOVA assumptions (4)

Answer 2

Absence of multicollinearity = correlation of independent variables to each other

Normal distribution of residuals (errors or how far away line is from your error)
- output in SPSS can be plotted to assess normality

Homogeneity of residual variance
- Levene’s test for homogeneity of variance

Linear relationship between covariates and outcome variable at each level of independent variable

Question 3

ANCOVA outputs

Answer 3

Least squared mean (LSM) or estimated marginal means = mean output which has been adjusted for covariates

Partial eta = effect size, above .15 —> considered good

Beta coefficient = is slope significantly different from 0 (strength of relationship between variables + covariates)

Question 4

What are dummy variables and how are they used?

Answer 4

Dummy variables - method of dichotomizing variables, independent continuous or multi-categorical variables which must be split into 2 categories for ANCOVA

Base case is ignored (000), then other groups coded using 0 or 1

Number of dummy variables is groups - 1

Used because regressions can’t have categorical variables with more than 2 groups (ex. 4 categories of physical activity –> med + high with base case low)

Question 5

Correlation

Types

What to report

Answer 5

measures strength of relationship between 2 continuous variables with a line of best fit through data

Pearson = parametric
Spearman = nonparametric

df = n - 2 (2 parameters), r vaue, p value

Question 6

Qualities of correlations to measure effect

Answer 6

Direction: positive or inverse

Strength: coefficient of correlation (r) or distance of observed data from line
r =between 0 to ±1, ±1 = perfect correlation, 0 = no correlation
Slope ≠ r value

Question 7

Strength brackets for correlation

Answer 7

Strength brackets (can depend on model used):

Small strength = 0.1-0.3
Medium strength = >0.3 - 0.5
Strong strength = >0.5 - 1.0

Question 8

Assumptions for correlations

Answer 8

Variables are continuous (no categorical data)

Variables are approximately normally distributed → otherwise use Spearman’s

Linear relationship between variables

No extreme outliers - they drastically change r value

Question 9

Association vs causation

Answer 9

correlation/regressions assess relatedness/association and NOT causation due to error and confounding variables

Question 10

Linear regressions

Answer 10

assesses relationships between 2+ continuous variables