Non-Parametric Group Comparisons Flashcards
What are Parametric Statistics?
- Assume data follow a normal distribution
- Make inferences about data parameters
What are the advantages of Parametric Statistics?
- More accurate and precise
- More statistical power, less Type II Errors
What are the disadvantages of Parametric Statistics?
- Not very robust, cannot be used if assumptions are violated
What are Non-Parametric Statistics?
- Do NOT make assumptions about normal distributions
- Do NOT make inferences about data parameters
What are the advantages of Non-Parametric Statistics?
- Simple
- More robust = Heterogeneity of Variance
What are the disadvantages of Non-Parametric Statistics?
- Less statistical power
- Not well suited for numeric interpretation
For interval or ratio, continuous data what type applications would be used?
Parametric Stats
What tests are considered Parametric Statistics?
- Paired T-Test
- Independent T-Test
- ANOVA
For nominal or ordinal, discrete data what type applications would be used?
Non-Parametric Stats
What tests are considered Non-Parametric Statistics?
- Wilcoxon Signed Rank Test
- McNemar Test
- Fisher Exact Test
- Wilcoxon Mann Whitney Test
- Kruskal-Wallis Test
- Chi-Square Tests
If data is interval or ratio that demonstrates NOT normal distribution or have heterogeneous variance, what type of statistics should be used?
Non-Parametric Statistics
If continuous data in a Paired T Test does not meet parametric assumption, what test should you shift to?
Wilcoxon Sign Test
If continuous data in an Independent T Test does not meet parametric assumption, what test should you shift to?
Wilcoxon Mann Whitney Test
If continuous data in an ANOVA Test does not meet parametric assumption, what test should you shift to?
Kruskal Wallis Test
What are the 3 ways to check of the data’s distribution is normal?
- Kolmogrove-Smirnov or Shapiro-Walk Tests
- Skewness
- Mean-Median-Mode
For Skewness what are the ranges for normal vs skewed?
- Between -1 to +1 = moderately normal
- <-1 or >+1 = skewed
If the continuous data innervation or ratio distribution deviates from normal based on the tests, what should be done?
Use the equivalent Non-Parametric test
In a test of normality such as Kolmogorov-Smirnov or Shapiro Will test what do the hypotheses mean?
H0: normal distribution
HA: non-normal distribution
If p-value is less than 0.05 in a test of normality, what does that mean?
Reject the H0, data is non-normal in distribution = use Non-Parametric Test
In a test of normality such as Skewness what does a negatively skewed data mean?
Mode > Median > Mean
In a test of normality such as Skewness what does a positively skewed data mean?
Mean > Median > Mode
For skewness, how does negatively and positively skewed data fall on the chart?
Negatively = tail points towards negative numbers
Positively = tail points towards positive numbers
Non-Parametric tests for Comparing Groups, 2 DEPENDENT groups, which ones are used in ORDINAL Dependent/Outcome variables?
- Sign Test
- Wilcoxon Signed Ranked Test
Non-Parametric tests for Comparing Groups, 2 DEPENDENT groups, which ones are used in DICHOTOMOUS Dependent/Outcome variables?
McNemar Test
Non-Parametric tests for Comparing Groups, 2 INDEPENDENT groups, which ones are used in ORDINAL Dependent/Outcome variables?
Wilcoxon Mann-Whitney Test
Non-Parametric tests for Comparing Groups, 2 INDEPENDENT groups, which ones are used in DICHOTOMOUS Dependent/Ouctome variables?
- Chi-Square Test of Homogeneity or Independence
- Fisher Exact
Non-Parametric tests for Comparing Groups, >2 INDEPENDENT groups, which ones are used in ORDINAL Dependent/Outcome variables?
Kruskal-Wallis Test
Non-Parametric tests for Comparing Groups, >2 INDEPENDENT groups, which ones are used in DICHOTOMOUS Dependent/Outcome variables?
Chi-Square Test of Homogeneity or Independence
What is the Signed Test?
Used with DEPENDENT groups, comparison between baseline and post treatment
What type of data is used in a Signed Test?
Can be used for analysis of ordinal, interval, or ratio data BUT MUST BE converted to ORDINAL data options of +, no change, or -
The Signed Test looks at what?
- Test Direction of Changes
- Number WITH change is compared to number with OPPOSITE direction of or NO change
What would the H0 of a Signed Test be?
N (-signs) = N (+ or no change signs)
N= number of signs
What is a Wilcoxon Signed Rank Test?
Used with DEPENDENT groups, comparison b/t baseline and post, compare RANKINGS among 2 DEPENDENT groups
What does the Wilxocon Signed Rank Test show?
Test Direction and Magnitude of Changes
What is the H0 of a Wilcoxon Signed Rank Test?
Group Median Pre = Group Median Post
Compares differences between GROUP MEDIANS
What is the data that is used in Wilcoxon Signed Rank Test?
Same as Signed Rank
Wilcoxon Signed Rank Test when compared to others is what?
- More POWERFUL than sign test
- More SENSITIVE than paired t-test when n<50
In a Wilcoxon Signed Rank Test, you compare smallest absolute value (T) to a critical value and what?
- n < 15 use Wilcoxon Critical Value (W)
- n > 15 use z-test
A positive and negative ranking in Wilcoxon Signed Rank means what?
+ = POST greater
- = PRE greater
What is the Wilcoxon Mann-Whitney Test?
Used with ORDINAL data, measures the rank patients from low to high
What is the H0 in a Wilcoxon Mann-Whitney Test?
No difference in the distribution of the scores in each group, no difference in the sum of ranks for each group
What is a Kruskal-Wallis Test?
ONE Way ANOVA
ORDINAL Data if there are more than 2 independent groups
What does the Kruskal-Wallis Test use for analysis?
MEAN RANKS instead of sum of ranks
What is the H0 of a Kruskal-Wallis Test?
Mean Rank a = Mean Rank b = Mean Rank c
If Kruskal Wallis has >5 subjects per group, then use what?
Use x^2 distribution with (k-1) degrees of freedom, where k is the number of groups
