Parametric and Non parametric Tests.pptx

comparasion between parametric and non parametric Stat Tests.

1. PARAMETRIC AND NON-PARAMETRIC TESTS Dr. Devesh Pandey MBBS, MD (Pharmacology) 8-12-2020 1

2. Statistical Tests Mathematical calculation or analysis of observed and collected data from the population or sampling. Statistical Power- Result obtained of tests, measure of probability that a statistical test rejects a false null hypothesis or probability of finding a significant result. Broadly, data classified into two- Qualitative data (non-numerical) Quantitative data (numerical) 2

3. DATA Smokers/non-smoker Hypertensive/ normotensive Mild/moderate/severe asthma Observation in numbers BP values Biochemical levels Countable, not ordered Male/female Yes/no Old/new Data in order/ scale Mild/moderate/severe Distinct/separate observation Blood group(O,A,B,AB) Gender(male/female) Values within finite/ infinite interval Ht., Wt., temp., RBS 3

4. Measure of Data (A) Mean- Arithmetic average of any given data. (B) Median- Middle value of any given data when it is arranged in ascending/ descending order. (C) Mode- Very frequently occuring event in given data. 4

5. TYPES OF DATA DISTRIBUTION (A) NORMAL ( Gussian ) DISTRIBUTION- Mean=Median=Mode Variables have tendency to cluster around central value with a symmetric distribution on either side. Skew is zero If tail skew towards right side- Positive skewed If tail skew towards left side- Negative skewed In skewed distribution Median- good measure of central tendency. Such made to normal distribution by suitable transformation by taking logarithum , square root or reciprocal. 5

6. 6

7. (B) BINOMIAL DISTRIBUTION- Data where only two possibilities. E.g. smokers/ non-smokers, male/ females, survived/ not survived. Two peaks 7

8. (C) POISSON DISTRIBUTION- Commonly used distribution in health sciences. Distribution of number of occurrences of some random event in an interval of time or space. 8

9. 9

10. 10

11. 11

12. Classification 12

13. Comparison between Statistical significance tests Parametric tests Non-Parametric tests Distribution Normal distribution Any Data set relationship Independent Any Assumed variance Homogenous Any Measure of central tendency Mean Median Type of Data Ratio/ Interval (Quantitative) Ordinal/ Nominal (Qualitative) Strength More powerful Less powerful Comparison Mean ±SD Percentage/ proportion 13

14. Equivalent Tests Parametric tests Non-Parametric tests Correlation test- Pearson Spearman rank order (Independent measures 2 groups) Unpaired student t-test Mann-Whitney U test (Repeated measures 2 condition) Paired t - test Wilcoxon signed rank test (Independent measures >2 conditions) ANOVA one way Kruskal-wallis test (Repeated measures > 2 conditions) ANOVA one way Friedman test 14

15. PARAMETRIC TESTS Parameter: is any numerical quantity that characterizes a given population or some aspect of it. P-value - (Statistical significance) Represents risk observed in experiment due to chance. Determines significance/non-significance of a result. <0.05 value- significant (uncertainty of 5%) Confidence interval (CI)- (Clinical significance) 95%- means 95% chance of cases to fall in the range. 15

16. Differences Between Means – Parametric Data t-Tests- compare the means of two parametric samples Excel: t-Test (paired and unpaired) – in Tools – Data Analysis E.g. Is there a difference in the mean height of men and women? E.g. A researcher compared the height of plants grown in high and low light levels. Use a T-test to determine whether there is a statistically significant difference in the heights of the two groups E.g. the length and weight of something –> parametric vs. did the bacteria grow or not grow –> non-parametric 16

17. ANOVA (Analysis of Variance)- compares the means of two or more parametric samples. Excel: ANOVA – check type under Tools – Data Analysis E.g. Is there a difference in the mean height of plants grown under red, green and blue light? E.g. A researcher fed pigs on four different foods. At the end of a month feeding, he weighed the pigs. Use an ANOVA test to determine if the different foods resulted in differences in growth of the pigs. 17

18. Advantages of non-parametric tests These tests are distribution free. Easier to calculate & less time consuming than parametric tests when sample size is small. Can be used with any type of data. Many non-parametric methods make it possible to work with very small samples, particularly helpful in collecting pilot study data or medical researcher working with a rare disease. 18

19. 19