SPRING ES, confidence intervals and NHST Flashcards
what can small es be sig
large samples
when can large es be sig
small samples
what are es
measure strength between two variables independent of sample size and measurement unit
standardised estimate of the difference
how to estimate es
previous research
pilots
guess in reference to study es (ie cohens)
cohens d es
d = diff between means / SD (control group or av of both)
how to calc a priori power
es want to achieve
sample size
power (min 0.8)
sig (
what to do if power too low
> sample size
OR make one tailed (but dodgy)
maximising es
es increase if random variation decreases (sd decrease = d increase)
use well controlled procedures to minimise confounds and possible variance
within > between BUT within harder to calc
n2 es
explained variability/total variability
between subj SD (SS) / total variability (SSt)
OR if not known
n2 = (g-1)/(g-1)f+g(n-1)
g= number of groups
f2 es
r1/1-r2
NHST
null hypothesis sig testing
probability that the null hyp is true
what happens if 95% CI falls within range to include 0
report no sig
no diff between groups
how can you interpret significance
inconsistent
equivocal
consistent
inconsistent sig
conflict in results and further research required
equivocal sig
One is statistically significant, the other not, but both are in the same direction. Therefore results are equivocal, more research is required.
consistent sig
The two results are in the same direction, & size of mean difference is similar. Therefore they reinforce each other and considered together, provide fairly strong evidence for the effect.
what is NHST
compare two hyp about what will happen if introduce exp manipulation or randomly sample group and measure a DV
hybrid of sig as esitmation of strength and alt/null using predetermined alpha
what is the p value
probability of obtaining an effect greater than or equal to that observed presuming that the null is true
