Chapter 14: Factorial Designs

Question 1

Factorial Design (ANOVA)

Answer 1

when an experiment has two or more categorical independent variables

used to compare the means of different groups to make judgments about the effects of the different IVs and possible interactions b/w the IVs

linear model

Question 2

independent factorial design

Answer 2

several independent variables and each has been measured using different entities (between group)

every group consists of different people

compare means

used when you expect a manipulation to have a long lasting effect

Question 3

repeated measures design

Answer 3

several independent variables have been measured, but the same entities have been used in all conditions

everyone exposed to every condition

more powerful since it takes away individual differences when everyone is their own control

Question 4

mixed design

Answer 4

several independent variables have been measured, with different entities, whereas others used the same entities

one factor is repeated measures and one factor is independent

Question 5

why use factorial designs?

Answer 5

examine the influence of 2 independent variables simultaneously: like in the real world which is complex and multifaceted

greater statistical power: if both variables don’t correlate highly and both actually do relate to the DV, for any 1 IV, you are gonna have a greater ability to find the effect that exists

test for interactions: testing for main effects does not allow you to catch everything going on

Question 6

main effect

Answer 6

statistical significance between the means (column/row)

Question 7

interaction

Answer 7

observed (cell) means are different than what you can expect if you were to just add the effect of factor a and the effect of factor b

Question 8

what is total variability (SSt) made up of ?

Answer 8

variance explained by the model/experimental manipulation (SSm) and residual/unexplained variance (SSr)

Question 9

what is the model variability (SSm) made up of ?

Answer 9

variable A variance (SSa), variable B variance (SSb), and variance explained by interaction of AxB (SSab)

systematic variance attributed to the model

Question 10

why do we use mean squares instead of sum squares?

Answer 10

mean squares adjusts for DF to undo the bias of SS. SS is affected by the number of people / groups

the bigger the MS, the bigger the F statistics, meaning that the null us much less tenable)

Question 11

why are follow up analyses needed?

Answer 11

main effects for 3 or more groups… which groups are different?

interactions… what type are they?

Question 12

the follow up approach you should use depends on…

Answer 12

1. whether you have an interaction or not
2. whether you have hypotheses on which groups are different
3. sample size
4. if you want a liberal/conservative p value
5. type of error that is most important for you not to make

• if there is an interaction present, do not interpret the main effects, even if they’re statistically significant
  *ONLY interpret main effects if the interaction is trivial

Question 13

graph + simple effects analysis

Answer 13

used when a significant interaction is present and you need to describe that interaction

looks at the effect of one IV at individual levels of the other IV

Question 14

graph + planned comparisons

Answer 14

used when there is no interaction, but there are a priori hypotheses about which groups are different

restricts the number of tests, so FW alpha stays less than .05 and there is greater power

Question 15

graph + post hoc tests

Answer 15

there is no interaction and you’re unsure which groups will differ. tests for sig main effects (if they are stat sig main effect, then do post hocs)

Question 16

how to tell if there in an interaction in graphs?

Answer 16

interaction is present if the lines are not parallel (they cross)

interaction is not present if the lines are parallel (do not cross), this means there are only main effects

in bar charts, there is an interaction if there is no consistency in differences across levels

Question 17

what is a median split?

Answer 17

A Median Split is one method for turning a continuous variable into a categorical one. Essentially, the idea is to find the median of the continuous variable. Any value below the median is put in the category “Low” and every value above it is labeled “High.”

it throws away a lot of data and gets rid of levels/continuum, decreases power

NEVER USE. just use a regression since it can handle continuous and categorical variables, whereas ANOVAs can only handle categorical