Nonparametric Tests Flashcards
(29 cards)
Parametric test
Knowledge of probability distribution
Nonparametric Test
- distribution free statistics / rank method
- doesn’t make assumption
- ranking or categorizing the data
When to use non parametric test
- small sample size
- not normally distributed, homoscedasticity
- outliers
- ordinal scale
Advantages
Easy to compute
Doesn’t need requirements of normality and homogeneity
Not affected by outliers
Disadvantage
Raw data, ranking
Risk of Type 2 error
1 group
One sample median
Chi square of goodness of fit
1 IV with 2 levels (paired)
Wilcoxon Signed-Rank
Sign
McNemar
1 IV 2 levels (independent)
Mann-Whitney
Median
Chi Square
Fischer’s Exact
1 IV 3 or more levels
Kruskal-Wallis
Chi square
Friedman 2-way ANOVA
Relationship between 2 variables
Spearman’s correlation
Ranking of data
1 assigned to smallest
N - largest
One sample median test
1 group
1 IV with 2 levels (paired)
Aka sign test
DV : ordinal
Measure if sample median differ from hypothesized value
Mann-Whitney U test
1 IV 2 levels (independent)
aka Willcoxon-Rank Sum Test NP: independent t-test n : at least 8 IV : nominal, dichotomous DV: OIR
Median Test
1 IV 2 levels (independent)
NP: k-sample test
Test equality of median
OI data
Sign Test
1 IV 2 levels (paired)
Alternative : paired t-test
(Single group)
- Determine if a population median = particular value
(Paired)
- median difference = 0
Could be used in place of Wilcoxon Signed Rank
- not ordinal, could be merely classified as positive or negative
Wilcoxon Signed Rank Test
1 IV 2 levels (paired)
aka Wicoxon Matched-Pairs Signed-Rank Test NP: paired ttest compares medians of 2 related groups More powerful than Sign test - take into account sign + magnitude of At least 5
Chi Square Test
1 IV 2 levels (independent)
3 or more levels
aka Pearson’s Chi Square
Relationship between 2 categorical variables
Analyze difference between observed frequency &; expected frequency
Z test
Parameter : proportion or percents
Predictor : one or 2 levels
Chi- square test
Parameter: frequency
Predictor : greater than 2 levels
Types of Chi-Square Test
- Goodness-of-fit
- Homogeneity / Independence
- Association
Chi-Square Goodness of Fit
Data FITS distribution
One categorical variable
Chi-square Homogeneity/Independence
Two/more IV
If diff samples come from populations with the distribution
Chi-square Association
Relationship or association between 2 QV
Computationally same as test of homogeneity except for E
Chi-square Goodness of fit
Observed set data is distributed as expected
Sample distribution is representative of the population
