Proportional Odds Model: Overview
Ivy Liu, Bhramar Mukherjee
Abstract
Ivy Liu, Bhramar Mukherjee
Abstract
Abstract Outcomes recorded on an ordinal scale, such as severity of migraine, stages of progression of disease, or degree of relief caused by treatment, occur commonly in clinical trials. Questionnaire data, which are often collected at each visit to gauge patient condition, routinely contain responses on an ordinal scale. Naïve dichotomization of the full ordinal scale leads to loss of information and efficiency while analyzing such outcomes. The proportional odds model is a popular choice for analyzing such ordinal outcomes. The current article provides a concise introduction to the proportional odds model and proceeds to review some more recent work related to this model. Specifically, we discuss the problem of analyzing stratified ordinal data in which observations from the same strata are correlated. Such correlated ordinal data may develop in multicenter clinical trials or with ordinal responses measured repeatedly on the same subject in longitudinal studies. Certain computing tools for fitting such models are discussed.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Abstract Outcomes recorded on an ordinal scale, such as severity of migraine, stages of progression of disease, or degree of relief caused by treatment, occur commonly in clinical trials. Questionnaire data, which are often collected at each visit to gauge patient condition, routinely contain responses on an ordinal scale. Naïve dichotomization of the full ordinal scale leads to loss of information and efficiency while analyzing such outcomes. The proportional odds model is a popular choice for analyzing such ordinal outcomes. The current article provides a concise introduction to the proportional odds model and proceeds to review some more recent work related to this model. Specifically, we discuss the problem of analyzing stratified ordinal data in which observations from the same strata are correlated. Such correlated ordinal data may develop in multicenter clinical trials or with ordinal responses measured repeatedly on the same subject in longitudinal studies. Certain computing tools for fitting such models are discussed.
Key concepts: Ordinal data, Ordinal regression, Ordinal Scale, Odds, Ordered logit, Econometrics, Scale (ratio), Odds ratio