week 7 Flashcards
What do we want to find with the model
simplest model which best explains the variance in our data
What do we need to know to choose which test
Independent/paired sample
Does our data meet the assumptions for parametric test: type of data, normality, homogeneity of variance, independence
What types test is there
Paired-samples t-test
Wilcoxon test
independent samples t-test
Mann-Whitney U test
Are the groups independent
Using within/between subjects design
If within-no
if between-yes
what is type of dependent variable
many models are based on linear
for testing our data needs to be: interval, ratio or ordinal
What is assumption of normality
we need to have a normal distribution for the test
What is homogeneity of variance
For paired samples we do not assume equal variance
scores need to be roughly equal at different point on the predictor variable
What are then different ways to test normality
Plot the data-does it look normal
Look at Q-Q plots (or P-P plots)
use a statistical test to compare your distribution to normality
What is independence
the assumption that errors in the model of our data not related to each other
What does a paired sample t-test do
compare means from two groups
if the groups come from the same population we expect the means to be roughly equal. we expect large difference between samples to be rare
Our null hypothesis is that the group membership has no effect
t=observed difference between sample means / estimated
We are interested of the difference between our groups is due to the effect we are testing
For paired samples t-test the appropriate effect size is Cohen’s d
What is Wilcoxon signed-rank test
uses the direction of the difference and the rank of the difference
Effect size: Z/SQUARED-N
