comparing two means

parameter

makes assumption about a population using probability distribution

nonparametric stats

do not make assumptions about a population

assumptions of parametric tests

scale data (ratio or interval)
random sampling
equal variance-- roughly equal before starting
normality- data sampled form normal distribution

true or false- you should toss outliers in a parametric test

false

what is a t-test

real difference v sampling error between two means

two levels of 1 independent variable

variance comes from two sources:

1- IV
2- everything else (error variance)

comparing means- independent groups

difference between means (including treatment effects and error) / variability within groups (difference in error alone)

error is all sources of variability that cannot be explained by IV

comparing means- repeated measures

difference between pairs (including treatment effects and error) / standard error of difference of score

t test score interpretation

t = difference between means / variability within groups

if t > 1- greater difference between groups
if t < 1- more variability within groups

independent t-test

A statistical test to determine whether there are significant differences between two independent groups' means being tested on the same dependent variable- is there a difference between groups?

independent t-test formula

difference between group means / variance within groups

assumptions for unpaired t-test

data is ratio or interval
samples are randomly drawn
homogeneity of variance (calculated via Levene's test)
population is normally distributed

what is effect size

effect that the IV has on DV
Cohen's d (or standardized mean difference)

Cohen's d sizes

Small: .2
Medium: .5
Large: .8

paired t-test

a statistical test to determine if there is a difference between means via repeated measures- is there a difference between conditions in the same person?

paired t-test formula

mean of paired difference scores / standard error of difference scores

assumptions for paired t-test

data is ratio or interval
samples are randomly drawm
no variance due to same participants
population is normally distributed

use of multiple t-tests will increase chance of making a _____________

type I error

ex- comparing 3 levels of same IV changes p value from .05 to .15

MDC v MDIC

MDC first