STAT 111 Lecture Notes - Lecture 17: Null Hypothesis, Test Statistic, Design Patterns

• School: University of Vermont
• Department: Statistics
• Course: STAT 111
• Professor: Jacob William Martin

Document Summary

Difference between what is observed and what is expected. Sample size: ecpcted counts for each cell > 5. If all the conditions are met, the test statistic follows a x^2 distribution. A chi-squared distribution will always be right skewed, and starting at zero. Family (like t) df are based on number of categories (or cells) Put in as right tailed on statkey. 1 categorical with 2 outcomes = single proportion. 2 categorical with 2 outcomes each = difference in proportions. 1 categorical with 3+ outcomes = x^2 - gof. 2 categorical with 3+ outcomes = x^2 = toa. Null hypothesis specifies proportions, p, for each of k categories of single categorical variable. Data = observed counts in each category. Compare expected counts = (n x p) for each cell to observed counts using. Find a p-value using the upper (right) tail of a x^2 distribution with k-1 df. Tests for association between two categorical variables.