Skip to content

Interactive and online course for analyzing cohort datasets

Notifications You must be signed in to change notification settings

lwjohnst86/analyzing-cohorts-course

Repository files navigation

Analyzing Cohort Data in R by Luke Johnston

This repository contains the files used by DataCamp to generate their online and interactive course content.

Learner persona

Link to student profiles. This course will be useful to two learners, and possibly a third (dependent on field of study):

  • Catalina: She will likely already know most of the basic statistical techniques in the course. However, this course may give her material on creating her own course, as this course will present various types of cohort study designs and how to interpret the results from the statistical analyses (in the cohort context). This will be useful to her medical students, who rely on this type of data to make medical decisions that could save people's lives.
  • Jasmine: The material in this course will teach her some fundamentals on how to apply statistical techniques using code. If she teaches health insurance policy research at her alma mater, knowing the specifics of analyzing cohort datasets would help make her students better with creating/analyzing/studying health insurance policies.
  • Mohan (or maybe another biomedical graduate student?): He will likely not be interested in the statistical portion of this course, though he may find the sections with the code to analyze and interpret the cohort data useful. If he was in any biomedical or health sciences graduate programs, he would find this course very useful.

General course outline

  • Chapter 1: Introduction to cohorts and types of research questions
    • Lesson 1: Introduction to cohort designs
    • Lesson 2: Introduction to the datasets and exploring them
    • Lesson 3: Scientific questions that can be asked of cohort data
  • Chapter 2: Data exploring, wrangling, and formatting
    • Lesson 1: Pre-wrangling exploration
    • Lesson 2: Tidy cohort data and wrangling into analyzable form.
    • Lesson 3: Transforming and modifying variables (maybe into two lessons?)
  • Chapter 3: Running the statistical techniques
    • Lesson 1: Common statistical techniques used for analyzing cohorts
    • Lesson 2: Adjustment, confounding, and modelling.
    • Lesson 3: Interaction testing and sensitivity analyses
    • Lesson 4: Extracting relevant data from results and post-modelling-wrangling {{wording needs changes}}
  • Chapter 4: Presentation of results from cohort studies
    • Lesson 1: Language and information to use when communicating the findings
    • Lesson 2: Back transforming results for easier interpretation
    • Lesson 3: Communicating your results through graphs
    • Lesson 4: Using tables to convey your findings

The dataset used is:

  • framingham: Framingham Heart Study.

Course Description

Cohort studies are a powerful study design that allows researchers to better understand how to reduce, manage, or prevent disease in a population. The findings from cohorts are critical to creating effective public health interventions. Because peoples' lives and health are directly impacted by the findings from cohort studies, there is immense pressure to ensure that the analysis done correctly and appropriately and that the presentation of results is as meaningful and as simple as possible. The difficulty in the analysis also comes from the wide range in study designs, data collection and type, and research questions. In this course, we'll be covering how and what to analytically ask of cohort data, what are special considerations for data processing, which statistical techniques to choose, and how to present the results for effective communication to public health professionals.

Learning Objectives

Overall

  • Learn how cohort data looks like and what questions to ask of it
  • Learn what to look for when exploring and pre-processing the data before the main analysis that may be more specific to cohort datasets
  • Learn how to think about and approach processing, analyzing, and presenting results from cohort studies in a way that is meaningful and interpretable to end users, the clinicians and the public health professionals

Chapter 1

  • Know that analyzing cohorts is just as much about the process/workflow as it is about the specific tools/statistics to use.
  • Recognize the type of cohort design from the data
  • Know which research questions can and cannot be asked of the given data
  • Have a general understand of which statistical tests to use for which data and that the test will dictate how the data should look.
  • Know where to look and what resources to use if you can't find the statistical or analytical techniques needed to analyze the data.

Chapter 2

  • How to get data into form for quick exploration
  • How to wrangle the data to get into a form to use for the analysis/statistic
  • Know the various tools to transform the data and when is more appropriate to use them
    • Know to be very careful of ever converting continuous into discrete
    • Know when and why to transform variables (good reasons and bad)
    • Reducing number of categories in categorical variable
  • Dealing with outliers (or not)

Chapter 3

  • Learn to run some common statistic techniques on different datasets and research questions
  • Learn how to extract the relevant results from these stats methods
  • Understand why its important to not use or even look at p-values in the output
  • Know how to understand and interpret the results (we'll get to knowing what exactly is most useful to present and show for higher impact)
  • Importance of examining multiple levels of adjustment (unadjusted, minimally, fully), and how this can inform what is going on in the relationship.
  • Knowing how to choose/decide on what variables to adjust for

Chapter 4

  • Give careful thought into how to best show results in a understandable and impactful way for public health professionals and clinicians.
  • Emphasize figures over tables, but know when to use either.
  • Be cautious in interpreting "non-significant" findings, since they could also have meaningful clinical implications.
  • Use cautious language when presenting results
    • Especially with "causal" language
  • Discuss STROBE guidelines in context of longitudinal study.

Prerequisites

About

Interactive and online course for analyzing cohort datasets

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published