1
2	Common Probability Models Characterization of a Probability Distribution and Summary Measure Type of random variable continuous, discrete, censored Summary measure used for outcome mean, geometric mean, proportion, odds, hazard Measure used for comparison of groups difference, ratio Quantification of statistical information
3	Common Probability Models Comparing means (e.g., Normal probability model of mean response) Continuously distributed outcome: e.g., blood pressure Outcome summarized by mean response Compare groups by difference in means Information measured by within group variance
4	Common Probability Models Comparing geometric means (e.g., Lognormal probability model-- log outcome is normal) Continuously distributed, skewed outcome: e.g., serum cholesterol, PSA Outcome summarized by (log) geometric mean (median) response Compare groups by (log) ratio of geometric means (medians) Information measured by within group variance of log transformed response
5	Common Probability Models Comparing binomial proportions Binary (dichotomous) outcome: e.g., tumor response, 30 day mortality Outcome summarized by probability of event Compare groups by difference in proportions Information from mean variance relationship under null or alternative
6	Common Probability Models Comparing binomial odds Binary (dichotomous) outcome: e.g., tumor response, 30 day mortality Outcome summarized by (log) odds of event Compare groups by (log) ratio of odds Information from mean variance relationship under null or alternative
7	Common Probability Models Comparing rates (Poisson probability model) Outcome counts events: e.g., number of lesions, number of infections Outcome summarized by (log) event rate Compare groups by (log) ratio of event rates Information from mean variance relationship under null or alternative
8	Common Probability Models Hazard ratios (Proportional hazards survival model) Right censored time to event: e.g., time to death Outcome summarized by hazard (semi-parametric) Compare groups by (log) ratio of hazards Information proportional to number of events