Tests for nominal data Flashcards
Nominal Data
- categories, counts/ frequencies
- e.g-number of yes/no responses
Expected Frequency (chance)
- Coin flip = 50/50 chance of heads or tails
- chance level = 0.5
-Rolling Dice = 1/6 of each number - chance of rolling 6 = 0.17
- is often 50/50
Binomial Test
- Is the proportion in MY data what l would expect by chance
- Tells US probability of finding observed proportion given expected proportion (chance)
Assumptions of binomial test
- Nominal data
- single dichotomy (1 variable with 2 exclusive outcomes - yes/no)
- Random sample
- independent data
- known expected distribution (chance)
Reporting binomial test
A binomial test revealed that this proportion is/isn’t significantly different than would be expected by chance (p = -)
chi- square Test of Independence
- one sample VS another
- nominal data
- Is frequency different in group A vs B
- yes/no and males/females
Assumptions of chi- square Test of Independence
- Nominal data
- 2 dichotomies male/female and yes/no
- Random sample
- independent dated
- sample of at least 40
- Expected N of 5 + in each category (frequency/chance)
chi-square test of independence Equation
= sum observed- expected)* 2)/ expected for each cell
- Add chi-square for each cell
- Need to look up p value in table to see if significant
- degrees of freedom (rowS-1) x (columns-1)
- If chi-square is the higher than critical + sig at alpha = sig difference
Reporting chi- square test of independence
- chi-square test of independence revealed the proportions were sig different to those expected by chance
violating assumptions of chi-square test of independence
- If under the table it says 1/2 cells have expected count less than 5 = Assumption not met
- Fishers Exact test is non-parametric alternative
- Look at table + ColMn of fishers exact
- A 2-tailed fishers exact teSt shows …
chi-square goodness-of-fit test
- How does an observed proportion differ from expected
- similar to binomial test
- more than 2 options
chi-square goodness-of-fit assumptions
- Nominal data
-multiple levels of a single dependent variable(e.g. location: Africa, Asia, Europe etc) - random sample
- independent score
- each category has expected N of 5 +
- DF= number of options -1
Reporting chi-square goodness-of-fit results
- chi-square = higher than critical - = significant at alpha
- There is a sig difference between observed VS expected proportion
One tailed Hypothesis
- Directional
- Predicts direction of effect
TwO tailed Hypothesis
- predicts a difference
- Doesn’t say direction
- Always make hypothesis but report 2 tailed p valve
