Savoy


Total Posts: 1 
Joined: Jan 2016 


I have the option of taking a statistics class next semester, in addition to my more rigorous core mathematics class on probability. However, I'm unsure if the content is useful for application in quantitative finance and machine learning, which are my areas of interest. If the content is not useful, then I can skip it in favour of a later class which focuses on other aspects of statistics.
I would appreciate it if people could please inform me as to whether I should be taking this class, so that I don't end up wasting time and money on something that has low relevance.
The lecturer gives the following description of the class:
Many realworld problems involve analysing data sets that are not normally distributed. For example, binomial data in the form of presence/absence recordings, Poisson data measured as counts of rare events such as car accidents, Gamma data for measurements of rainfall and Weibull data for the expected lifetimes of machinery. This unit provides experience in analysing such observations. The majority of the unit concentrates on the presentation and analysis of such data sets. Generalised Linear Models (GLMs) are used to incorporate explanatory variables into the analyses. In developing these skills students are trained in an appropriate statistical software package. The unit also provides a rudimentary understanding of probability and statistics necessary for applying the likelihood theory for estimating these models.
The content of the class is as follows:
Bernoulli & binomial distributions. Statistical inference for proportions: hypothesis testing and confidence intervals.
Chisquared test and statistical inference for ratios: univariate hypothesis testing and confidence intervals.
Introduction to sampling schemes for the collection of data. Basic tools for analysing count data: Fisher’s Exact and CochranMantelHaenszel tests.
Logistic regression for binary response data.
Logistic regression for binary counts and conditional logistic regression for matched casecontrol data.
Poisson (loglinear) regression for Poisson counts.
Negative binomial regression for overdisperse count data.
Introduction to likelihood functions – construction and visualisation.
Maximum likelihood estimation (MLE) – analytic and numeric solutions.
Sampling distribution of MLE.
Properties of MLE. Comparison of likelihood ratio test (LRT) & Wald's test. Model goodness of fit based on deviance and Akaike's Information Criteria (AIC).
Thanks everyone.



