Which Stats Method?

This course will provide a basic overview of general statistical methodology that can be useful in the areas of infectious diseases, environmental medicine, and labwork. By the end of this course, students will be able to identify appropriate statistical methods for a variety of circumstances.

This course will not teach students how to implement these statistical methods. The aim of this course is to enable the student to identify which methods are required for their study, allowing the student to identify their needs for subsequent methods courses, self-learning, or external help.

You should take this course if you are one of the following:

  • Have experience with applying statistical methods, but are sometimes confused or uncertain as to whether or not you have selected the correct method.
  • Do not have experience with applying statistical methods, and would like to get an overview over which methods are applicable for your projects so that you can then undertake further studies in these areas.

This course covers the following scenarios:

  • Identifying continuous, categorical, count, and censored variables.
  • Identifying exposure and outcome variables.
  • Identifying when t-tests (paired and unpaired) should be used.
  • Identifying when non-parametric t-test equivalents should be used.
  • Identifying when ANOVA should be used.
  • Identifying when linear regression should be used.
  • Identifying the similarities between t-tests, ANOVA, and regression.
  • Identifying when logistic regression models should be used.
  • Identifying when Poisson/negative binomial and cox regression models should be used.
  • Identifying when chi-squared/fisher’s exact test should be used.
  • Identifying when data does not have any dependencies (i.e. all observations are independent of each other) versus when data has complicated dependencies (i.e. longitudinal data, matched data, multiple cohorts).
  • Identifying when mixed effects regression models should be used.
  • Identifying when conditional logistic regression models should be used.