Complex outcomes
Background
This chapter offers tutorials on handling non-binary outcomes in data analysis. The first tutorial, on Polytomous and Ordinal outcomes, delves into multinomial and ordinal logistic regressions, teaching methods to analyze datasets with more than two categorical outcomes and how to assess the fit of ordinal logistic models. The second tutorial introduces Survival Analysis, guiding readers through the components of survival data, right censoring, the Kaplan-Meier method, the Cox regression model, and the complexities of analyzing time-dependent covariates and survey data. Lastly, the tutorial on Poisson and Negative Binomial focuses on regressions related to count data, showcasing both regular and survey-weighted methods, and highlights statistical techniques to understand factors influencing certain variables.
Datasets:
All of the datasets used in this tutorial can be accessed from this GitHub repository folder
Overview of tutorials
In the preceding chapters, we have primarily focused on binary or continuous outcomes. In this chapter, we will discuss about categorical [outcome variable is categorical with more than two categories (i.e., multicategory)], survival [time-to-event outcomes and aims to model the time until an event happens, accounting for censored data (observations where the event has not occurred by the end of the study)] and count data [non-negative integers (i.e., 0, 1, 2, 3, …)] outcomes.
Polytomous and ordinal
The tutorial covers methods for analyzing polytomous and ordinal outcomes in R. Initially, it introduces a function to exclude invalid responses from datasets. The data is loaded, and its structure is checked. The tutorial delves into the multinomial logistic regression, where tables are created to explore data characteristics. Two models are fitted: one for unweighted data and another for survey-weighted data. It then transitions into ordinal regression. Here, outcomes are ordered, and ordinal logistic models are fitted, again for both unweighted and survey-weighted data. The final part of the tutorial assesses the fit of the ordinal logistic model using various variables.
Survival analysis
The tutorial provides an in-depth understanding of survival analysis. It starts by introducing the lung dataset which contains various attributes related to patients’ health. The concept of censoring, specifically right censoring, is detailed, which happens when a subject leaves the study without experiencing the event of interest. The tutorial delves into the components of survival data, introducing the event indicator, survival function, and survival probability. It then transitions to practical applications, demonstrating how to create survival objects, estimate survival curves using the Kaplan-Meier method, and visualize these curves. There’s also an emphasis on understanding and estimating median survival time and comparing survival times between different groups using the log-rank test. The Cox regression model, which is pivotal in survival analysis, is elaborated, along with the idea of hazard ratios and assessing proportional hazards. Time-dependent covariates, which account for variables that may change over time, are also touched upon. Towards the end, the tutorial addresses how to perform survival analysis on complex survey data, covering design creation, Kaplan-Meier plotting, and the Cox Proportional Hazards model in the context of these surveys.
Survival analysis in NHANES
The tutorial demonstrates an analysis of NHANES data linked with mortality data, and conducts a survival analysis.
Poisson and negative binomial
The tutorial provides a comprehensive guide on statistical analysis techniques related to Poisson and negative binomial regressions. Initially, it prepares the data by converting them into appropriate formats. A weighted summary of the dataset is then generated, emphasizing the distribution of a variable. The tutorial proceeds to demonstrate both regular and survey-weighted Poisson regressions, focusing on the relationship between fruit consumption and various factors. Finally, it explores negative binomial regressions and their survey-weighted counterparts.
Optional Content:
You’ll find that some sections conclude with an optional video walkthrough that demonstrates the code. Keep in mind that the content might have been updated since these videos were recorded. Watching these videos is optional.
Bug Report:
Fill out this form to report any issues with the tutorial.