Tests For Nominal Data Flashcards
Binomial test
Nominal data and want to know prob of finding results given expected results (chance)- expected often 50 50 but change based on target pop
Binomial test assumptions
Data are nominal, single dichotomy (one variable with 2 outcomes), scores are from a random sample, data are independent (ps contribute one data point), know the expected distribution of scores - can only be used on one sample
Diff between binomial and chi square test of independence
Binomial is one sample bs expected freq, chi is one sample vs another sample e.g. is freq of yes and no diff in group one than group 2
Chi square test of independence assumptions
Data are nominal, two dichotomies (male/female and yes/no), scores are from a random S pale , data are independent, sample of at least 40, each category must have N of 5 or above
How to calculate expected frequencies for chi square
The product of marginal totals. For each cell, row total x column total divided by overall total
How to calculate chi square
O-E, O-E squared, O-E squared divided by E. do for all cells then find the sum. Then look up p value and df in a chi square table . Df is rows-1 times columns-1. Chi is high- higher than critical then sig
How to report results of chi square
In the x condition, 50/100 reported feeling better while in placebo, only 15/100, a chi square test of independence revealed that these proportions were sig diff to those expected by chance (x^2(df)=a, p=)
Chi square test of independence non parametric alternative
Fishers exact test
Why we use two tailed results
One failed hyp is directional, two tailed is no directional. For a one tailed test to be sig, results need to fall in one extreme 5% of distribution so miss effect on the other tail. For 2 tailed, results have to fall in top or bottom 2.5% so scores must be more extreme to be signed
Chi square goodness of fit test
How does observed data differ from expected.
Chi square goodness of fit assumptions
Data is nominal, multiple levels of single DV (location), scores are from a random smpale, scores are independent, each category has expected n of 5 or above - calc like normal chi square?
