Odds ratios from logistic, geometric, Poisson, and negative binomial regression models

Abstract

BackgroundThe odds ratio (OR) is used as an important metric of comparison of two or more groups in many biomedical applications when the data measure the presence or absence of an event or represent the frequency of its occurrence. In the latter case, researchers often dichotomize the count data into binary form and apply the well-known logistic regression technique to estimate the OR. In the process of dichotomizing the data, however, information is lost about the underlying counts which can reduce the precision of inferences on the OR.MethodsWe propose analyzing the count data directly using regression models with the log odds link function. With this approach, the parameter estimates in the model have the exact same interpretation as in a logistic regression of the dichotomized data, yielding comparable estimates of the OR. We prove analytically, using the Fisher information matrix, that our approach produces more precise estimates of the OR than logistic regression of the dichotomized data. We also show the gains in precision using simulation studies and real-world datasets. We focus on three related distributions for count data: geometric, Poisson, and negative binomial.ResultsIn simulation studies, confidence intervals for the OR were 56–65% as wide (geometric model), 75–79% as wide (Poisson model), and 61–69% as wide (negative binomial model) as the corresponding interval from a logistic regression produced by dichotomizing the data. When we analyzed existing datasets using our approach, we found that confidence intervals for the OR could be up to 64% shorter (36% as wide) compared to if the data had been dichotomized and analyzed using logistic regression.ConclusionsMore precise estimates of the OR can be obtained directly from the count data by using the log odds link function. This analytic approach is easy to implement in software packages that are capable of fitting generalized linear models or of maximizing user-defined likelihood functions.Electronic supplementary materialThe online version of this article (10.1186/s12874-018-0568-9) contains supplementary material, which is available to authorized users.

Grappling With the COVID-19 Health Crisis: Content Analysis of Communication Strategies and Their Effects on Public Engagement on Social Media

“…As for RQ2, which involved examining the relationship between the 12 subdimensions (independent variables) and the public’s responses in terms of the count number of shares, comments, and likes (dependent variables), we first employed Poisson regression, a count regression model in SPSS [ 78 , 79 ]. However, real-world data sets are commonly known to violate the assumption in the Poisson regression with respect to overdispersion of outcome variables [ 80 ]. As expected, such a violation was detected in our data set, and thus, we followed the common practice of replacing Poisson regression with the NB2 [ 80 ] to improve the goodness of fit, especially Akaike information criterion and bayesian information criterion.…”

Exploring drivers for public engagement in social media communication with medical social influencers in China

“…First, we employed SPSS to examine the mean and variance of the count outcomes to determine if there was any violation of assumption in Poisson regression. It is not surprising to uncover a violation of a major assumption in Poisson regression in the real-world dataset [67] involving an overdispersion of outcome variables. In order to improve the goodness of fit of the model, we replaced Poisson regression with Negative binomial regression (NB2) [66].…”

“…Concerning collectivity, the results revealed that complimenting others in MSI communication was used significantly less than quoting others' posts/making reference to others' posts and asking questions. Users are motivated to seek health-related information for a purpose such as resolving a health issue [67], and hence, the use of overly polite expressions such as complimenting others might be inappropriate in this context. Asking questions was linked to public engagement in the form of a higher number of comments possibly because questions posed to MSIs answer a user's specific health problems but at the same time, promote dialogue and interaction among other users [4,70].…”

Efficacy of Diltiazem for the Control of Blood Pressure in Puerperal Patients with Severe Preeclampsia: A Randomized, Single-Blind, Controlled Trial

“…Additionally, we fitted a binomial negative model to estimate the average number of hypotension episodes on days of hospitalization and incidence rate ratio. After fitting the model, residual deviance was used to perform a chi-square goodness of fit test for the overall model [ 30 , 31 ].…”