Evaluating distributions: - Null hypothesis is typically two distributions are the same - Reject the null hypothesis = the two distributions are different - Threshold for p-value analysis typically around 5% and can range between 1-10% - Depends on the distributions being compared and what level is required to confirm difference
2 Sample Komogorov-Smirnov Test (aka: KS Test): - Determine if two datasets differ significantly. Used for continuous distribution - Does not make assumptions about distributions of data - Non-parametric & distribution free - D stat = max vertical distance between two distributions (smaller is better) - p value = low leads to believing the two distributions are different
Student's t-Test (t-Test): - Determine if two datasets are differ or from same distribution - Handles problems associated iwth inference based on "small" samples - Accomodate calculated mean could deviate from real mean - p value = low leads to supporting the two distributions are different