Regression Model for Count Data课件.pptx

Regression Models for Count DataChenYafei(陈亚飞)山东大学 公共卫生学院 生物统计学系Department of Biostatistics, School of Public HealthShandong University 2015.07.29OutlineCount Data DistributionPoisson RegressionNegative Binomial RegressionZero-inflated modelHurdle model Count DataA type of data in which the observations can take only the non-negative integer values {0, 1, 2, 3, ...}, and where these integers arise from counting rather than ranking.Number of events in fixed area and/or length of timeExampleNumber of hospitalizations over a period of timeSeizure counts for epileptics in two weeksNumber of new enhancing lesions seen on monthly MRI of the braincount dataBlood cells in a blood sample (space=volume) Number of patients “at risk” with a differing years of exposureNumber of bacteria in different areas of a Petri plate Number of trades in a time interval.Number of crimes on campus per semester.PoissonAll of themare this typeBinomialNegative binomialCount Data DistributionBinomial Distribution?Count Data Distribution Poisson Distribution???Count Data Distribution Poisson DistributionWe can come up with the Possion model by thinking of events counts as counts of rare events.aircraft accident, tumor, seizure counts in epileptics in certain time, hospital times in one year , traffic accidents in a cross in one month…不具有传染性、无永久免疫、无遗传性且发病率很低的疾病Sum of independent Poisson random variables is also PoissonCount Data DistributionNegative Binomial DistributionThe negative binomial distribution is based on an experiment satisfying the following conditions:1. The experiment consists of a sequence of independent and identical trials2. Each trial can result in either a success, S, or a Failure, F.3. The probability of success is constant from trial to trial, so P (S on trial i) = p for i = 1, 2, 3, …4. The experiment continues (trials are performed) until a total of r successes have been observed, where r is a specified positive integer.Count Data DistributionNegative Binomial Distribution