Loveday L Conquest
Overview of generalized linear models (GLMs), their use in forestry, fisheries, wildlife ecology, and environmental monitoring. Analysis of the statistical tests that fall under GLMS: chi-square tests on contingency tables, t-tests, analysis of variances, etc. Statistical software S+/R used throughout. Offered: Sp.
COURSE OBJECTIVE: The main objective of this course is to become familiar with a large class of statistical models known as Generalized Linear Models, or GLMs, and their use in forestry, fisheries, wildlife, ecology, and environmental monitoring. These statistical models form an umbrella under which fall many familiar statistical tests: chi-square tests on contingency tables, t-tests, analysis of variance, etc. Additionally, using GLMs allows one to drop the normal distribution assumption and substitute other error distributions, wuch as the binomial (e.g., mortality data), Poisson (e.g., animal or event count data), Weibull (e.g., survival data). Thus one can do parameter based testing and model building using distributions other than the normal. The statistical software R is used throughout the course. Examples from fisheries, forestry, wildlife, and other environmental or ecological scenarios are emphasized.
Student learning goals
Be able to identify linear models, generalized linear models, generalized additive models.
Be able to run these models and do model-checking along the way to move to a "best" final model.
Be aware of the limitations of these types of models.
General method of instruction
Class meets MWF 10:30-11:20; "R" computer labs are Friday 11:30-1:20, with formal lab instruction in the first hour and an "office hour" during the 2nd hour to learn "R" and to apply the techniques learned in class. Instruction in class will be traditional lecture format.
Prerequisites: concepts of likelihood ratio tests (STAT 512-13 or STAT 516-17 or STAT 341-342 or STAT 481), matrix algebra, calculus. QERM 514 is different from the QSCI 482-483-480-486 series in that it is NOT a "service course" and it is taught at the graduate level. It is a required course for 1st-year QERM students and is part of the QERM MS/Ph.D. qualifying exams each June. There is a prerequisite of one course in mathematical statistics, which can be fulfilled by having had STAT 512, or STAT 341-342, or STAT 481, Intro to Mathematical Statistics. In addition, knowledge of calculus and matrix algebra are required. It is assumed that the student knows the THEORY behind likelihood ratio tests; pages from a chapter in a standard mathematical statistics textbook can be made available. While the course is not overly theoretical, it definitely helps to be one who enjoys the ins and outs of mathematical statistics and likelihood ratio tests. A key concept that we will be making use of all quarter is, "-2 times the log of the likelihood ratio has an asymptotic chi-square distribution."
Class assignments and grading
There will be weekly homework assignments, many of which require use of the computer to analyze data sets. There will be a take-home final exam. For students in the QERM program, the take-home final exam acts as a precursor to the statistics portion of the QERM Applied Qualifying Exam.
Grades are assigned based on the weekly HW assignments, the project, the take-home final, and to some extent class participation.