t-tests between + within Flashcards
(13 cards)
If obtained t score is larger than t-table, probability of getting that large by chance is…
less than .05 P<.05
reject the null hypothesis
the independent variable is influencing something
Between-subjects t-test
ASSUMPTIONS
- dependent on interval or ratio scale
- both samples norm distr
- populations equal variance - homogeneity, levenes test
- Independence of observations
Between subjects t-test STEPS
- estimate popn SD (each group)
- pooled s2pool (n -1)
- Standard error of mean diffs. (n is whole number)
- mean differences
- t-score using what you calculated
- t-table, obtained larger than table reject null
- cohens d uses s2 pool
one-tailed
one
- double the alpha level .1 on table
- know before ‘a priori’ whether correlation is positive or negative
- null r = 0
- positive r > 0
- negative r < 0
two-tailed
two
- 0.05 on table
- cant predict/doesnt specify whether correlation + or -
- null = no correlation btwn variables
- alternative = is correlation
Outcomes - null or alternative hypothesis
H0 null = the groups show no performance difference
H1 = the groups show performance difference
bessels correction N-1
only for estimating, ie population standard deviation
T-Distribution is
the distribution obtained from an estimated population SD
t-score is
samples over or under estimate standard error?
how many standard error units our sample mean is away from the mean
Samples OVERestimate standard error
t is platy, increase sample =z-norm
effect size
Cohens D values
estimates proportion of variability in scores on dependent variable explainable by variation in the level of the indepdendent variable
- is result meaningful?
small .2, med .5 large .8
repeated measures (Within) is
experience both levels of condition (IV)
Repeated measures assumptions
- dependent variable on interval/ratio
- sample diff scores norm distrib.
- scores are related across conditions
–> same individual
–> less variability of individual differences
repeated-measures steps
- estimate variance of popn
- estimate standard error
- obtained t-value
- t-table (if obtained is larger, reject null)