Stat 301 – Day 9 Fisher’s Exact Test Quantitative Variables
Recap In analyzing two-way tables, the p-value tells us whether the difference in the group proportions/relative risk could have happened by the random assignment process alone Simulated the random assignment process to see whether our observed result was extreme “Fisher’s Exact Test”: Use counting methods to determine the exact probability
Investigation (p. 72) Only 6 of 21 minorities coached at third 24 nonminorities coached at third and 15 at first How set up two-way table? How define random variable?
Investigation Two-way table successes Group A p-value = P(X < 6) If we let X represent the number of minorities at third, want to find P(X < 6) Hypergeometric with N = 60, M = 30, n = 21 = C(30,6)C(30,15) + … =.0146 C(60, 21) failures
Investigation Two-way table successes Group A p-value = P(X < 6) If we let X represent the number of minorities at third, want to find P(X < 6) Hypergeometric with N = 60, M = 21, n = 30 = C(21,6)C(39,24) + … =.0146 C(60,30)
Quiz 6
Big Picture Comparing two groups on a categorical response variable Appropriate graphical summary (seg bar graph) Appropriate numerical summaries (conditional proportions, relative risk, odds ratio) Is the difference statistically significant? Fisher’s Exact Test: How often get a difference at least this large by the random assignment process alone Scope of conclusions Cause and effect? Generalize beyond those in study? Compare results Randomized?
Big Picture Do it all again! Compare groups on a quantitative response variable Graphical summaries Numerical summaries Statistical significance Scope of conclusions
Investigation (p. 102) Match the histogram with the variable (“Probability and Statistics for Engineers and Scientists”) Most important – your justifications
Stat 301 data
The moral: Try to anticipate variable behavior/explain patterns and deviations from patterns
Investigation 2.1.2
Aside: History of Statistics and Agriculture
Investigation (a) Experiment or observational study? Imposed seeding/unseeded Experimental units? clouds (b) Explanatory and response variable?
Investigation Center Spread Shape Unusual observations Always label!!!
Skip to Minitab detour (p. 110) Course Materials > ISCAM Data Page Minitab: Chapter 2, Minitab Files, Cloud Seeding.mtw Instructions in text R: Chapter 2, TXT files, Cloud Seeding.txt Handout Boxplots Dotplots Descriptive statistics
Graphical and numerical summaries Five number summary Median = ( )/2 = 44.2
Five number summary Unseeded Min=1.0 Q 1 = 24.4 median=44.2 Q 3 =163 Max= Seeded Min=4.1 Q 1 =92.4 median=221.6 Q 3 =430 max=2745.6
Boxplots IQR1.5IQR
Boxplots % of data lie above mean
For Wednesday Mini-project 1 proposal Finish Investigation through part (n) See online solutions, bring questions to class PP (p. 113) Combine parts (b) and (g) together (c)-(f) in Blackboard as multiple choice Investigation parts (a)-(d) (p )