Lecture 10

Question 1

omnibus effect

Answer 1

if a significant
difference exists across the dataset as a whole

Question 2

does random noise exist between columns and within columns?

Answer 2

In other words,
whatever “random noise” exists between columns also exists across rows (i.e., within columns)
in a completely random data set yes, the ratio is expected to be around 1

Question 3

what happens if we increase all data in a completely random data set one row by 2

Answer 3

variance between columns is now increase whereas the variance within columns is unchanged
values are no longer random

Question 4

Between column variability

Answer 4

how the columns differ from one another

Question 5

analysis of variance

Answer 5

ANOVA
statistical test where two or more groups can be
analyzed by taking the ratio of between group variance over within group variance

Question 5

Within column variability

Answer 5

how scores within each column vary from one another

Question 6

ANOVA ratios if H0 or H1

Answer 6

If the null is true, this ratio is expected to be = 1. If the null is false, this ratio is expected to be > 1
When the null hypothesis is false, the existence of an effect will alter at least one of the group
means, but the variance within that group will be unchanged if the transformation is linear (between-group variance will change, but the within-group variance will not)

Question 7

anova s^2 if H0 or H1

Answer 7

when null is true: s^2 between and s^2 within are expected to be equal
when null is false: s^2 between > s^2 within

Question 8

f ratio

Answer 8

between-to-within group variance ratio.
F distribution is derived from the t distribution. Specifically, it is the t distribution squared. The
mean of F = 1, and the mean of t = 0

Question 9

ANOVA assumtions

Answer 9

1. Data is normally distributed for all conditions
2. There is homogeneity of variance
3. There is independence of observations

Question 10

windsorizing

Answer 10

replacing extreme data points with less extreme values from the dataset, typically using percentiles

Question 11

t vs f

Answer 11

t and F are functionally equivalent. If a dataset with group groups is calculated as either a t test or
an ANOVA, the obtained F value will equal the t2, and the critical F value will equal the critical t
- but anova can do more than 2 groups

Question 12

source table

Answer 12

breaks up the model into its
individual source components
here, the between and within s2 (described as mean square, or MS)
are also split into their SS and df components

Question 13

anova and multiple regression relationship

Answer 13

ANOVA is to t what multiple regression is to Pearson’s r.

Question 14

partial correlation

Answer 14

completely control for the relationship

Question 15

semi-partial correlation

Answer 15

control for the
relationship for one of the variables)

Question 16

things to consider when having more than one predictor variable

Answer 16

1. these predictor variables may be related to one another. If they are correlated with one
   another, and thus not independent, it can affect the overall regression model. In this case, partial
   correlation and semi partial correlation can be used to piece apart the influence of overlap
2. it is important to ensure that all X variables are linearly related to the predictor variable, as
   well as linearly related to one another