# Statistical Reasoning in Public Health II – JHSPH 140.612.81

This 2009 JHSPH OpenCourseWare graduate course is the second half of a course providing a broad overview of biostatistical methods and concepts used in the public health sciences, emphasizing interpretation and concepts rather than calculations or mathematical details. It develops ability to read the scientific literature, to critically evaluate study designs and methods of data analysis, and it introduces basic concepts of statistical inference, including hypothesis testing, p-values, and confidence intervals.

Topics include comparisons of means and proportions; the normal distribution; regression and correlation; confounding; concepts of study design, including randomization, sample size, and power considerations; logistic regression; and an overview of some methods in survival analysis. The course draws examples of the use and abuse of statistical methods from the current biomedical literature.

Learning Objectives

Upon completion of this course, you will be able to:

• recognize different study designs and understand the pros and cons of each.
• learn methods for randomly assigning subjects to two groups.
• understand the concepts of confounding and statistical interaction; know how to recognize each.
• explain the relationship between power and sample size; use Stata to perform sample size calculations.
• create a scatterplot to visually assess the nature of an association between two continuous variables.
• interpret the calculated values of the correlation coefficient and the coefficient of determination, and understand the relationship between these two measures of association.
• perform a simple linear regression using Stata, and use the results to assess the magnitude and significance of the relationship between a continuous outcome variable and a continuous predictor variable, and for predicting values of the outcome variable.
• Understand why multiple regression techniques allow for the analysis of the relationship between an outcome and a predictor in the presence of confounding variables.
• perform a multiple linear regression using Stata, and use the results to assess the magnitude and significance of the relationship between a continuous outcome variable and multiple continuous and categorical predictor variables, and for predicting values of the outcome variable.
• perform a multiple logistic regression using Stata, and use the results to assess the magnitude and significance of the relationship between a dichotomous outcome variable and multiple continuous and categorical predictor variables.
• interpret the results from a proportional hazards regression model.

Course Organization

The content of this second of two courses is divided into eleven lectures spread out over four "modules."  The four modules are:

• Module 1: Issues in Design Study - Sessions 1–3
• Module 2: Linear Regression - Sessions 4–6
• Module 3: Logistic Regression - Sessions 7–9
• Module 4: Survival Analysis - Sessions 10 and 11

PDF slides and audio recordings from lectures are available for each session. Additionally, practice problems and homework assignments were fortunately included with the associated OpenCourseWare course. A final exam was given for this course, but it is not available through OCW.

No text was originally required for this course. However, several books were listed as "useful, but optional":

* Altman, D. G. (1990). Practical Statistics for Medical Research. Boca Raton, Florida: CRC Press.

* Freedman, D., R. Pisani, and R. Purves. (2007). Statistics. 4th Edition. New York: W. W. Norton & Company, Inc.

* Moore, D., G. McCabe, and B. Craig. (2012). Introduction to the Practice of Statistics. 7th Edition. New York: W. W. Norton & Company, Inc.

Recommended reading was assigned with some sessions, in all cases from Practical Statistics for Medical Research. Consult the Recommended Reading associated with each session.

Other Requirements

Students are also required to have access to Small Stata, a version of Stata that is less powerful (in terms of the amount of data it can store and process, not in terms of functionality) than regular Intercooled Stata, and costs significantly less. Small Stata carries a one-year users license. However, if you intend to further your study of statistics beyond this course, you may wish to purchase a copy of Intercooled Stata 8.

This flowchart should aid in choosing a correct statistical procedure during this course: Choosing a Correct Statistical Procedure

This document provides more information on performing a paired t-test in Stata: Paired T-test in Stata

Lessons Status
1

2

3

4

5

6

7

8

9

10

11