ANOVA Flashcards
types of t-test
independent samples t test
paired samples t test
one-sample t test
independent samples t test
compares means from 2 independent groups
paired sampled t test
compares means from 2 sets of individuals
repeated measures, matched subjects
one sample t test
compares observed mean to population mean
when to use t test over anova
more efficient with 2 groups
when to use anova over t test
more efficient with more than 2 groups
when to use anova
when want to compare more than 2 conditions
have 2 or more groups/conditions and more than one IV/factor
advantages of anova
can investigate effect of multiple factors on DV at same time
why not just use several t tests
this can increase chance of type 1 error - experiment wise/familywise error rate
anova controls for errors so type one errors remain at 5% so you can be confident significant results aren’t down to chance
anova assumptions
DV at interval or ratio level
Data from normally distributed population
Homogeneity of variance
For independent groups design, independent random samples taken from each population
nominal data
e.g. gender
numbers distinguish categories but no ranking
ordinal data
use scale to order/rank
size of number and differences mean nothing
interval data
scores in order, equal differences, no absolute 0
e.g. temperature
ratio data
e.g. height
scores in order, equal differences, absolute 0
check for normally distributed data
histogram
skew and kurtosis in descriptives table
what to do to skew and kurtosis values in descriptives table
convert to z scores by dividing by their std. error
> +/-9.96 then significant (p<.05) and suggests non-normal data
between-groups variance
variation between group means
from individual differences, treatment effects and random effects
within-groups variance
variation between people in same group
error variance
not from experiment
from individual differences and random effects
what is F
variance due to manipulation of factor error variance
how does anova calculate f ratio
due to manipulation of IV (BGV) and error variance (WGV) by dividing BGV/WGV
if error variance is smaller…
F=>1 and is significant
if effect of IV is smaller…
F=<1 and is not significant
p value must be what for it to be significant?
=/<.05
f ratio table
difference between F in anova and MR
MR predicts continuous outcome on basis of 1+ continuous predictor variables
ANOVA predicts continuous outcome on basis of 1+ categorical predictor variables
F ratio in both is the same but regression model for ANOVA contains categorical variables
IV
factor
factor
IV
levels of factors
conditions
conditions
levels of factors
mixed anova designs
1+ within subjects factors and 1+ between subjects factors
between subjects factors
vary between ppts
within subjects ppts
vary within ppts
describing anova designs
- how many factors in design?
- how many levels in each factor?
- whether factors are within/between subjects?