The Analysis of Count Data: A Gentle Introduction to Poisson Regression and Its Alternatives

Abstract

Count data reflect the number of occurrences of a behavior in a fixed period of time (e.g., number of aggressive acts by children during a playground period). In cases in which the outcome variable is a count with a low arithmetic mean (typically < 10), standard ordinary least squares regression may produce biased results. We provide an introduction to regression models that provide appropriate analyses for count data. We introduce standard Poisson regression with an example and discuss its interpretation. Two variants of Poisson regression, overdispersed Poisson regression and negative binomial regression, are introduced that may provide more optimal results when a key assumption of standard Poisson regression is violated. We also discuss the problems of excess zeros in which a subgroup of respondents who would never display the behavior are included in the sample and truncated zeros in which respondents who have a zero count are excluded by the sampling plan. We provide computer syntax for our illustrations in SAS and SPSS. The Poisson family of regression models provides improved and now easy to implement analyses of count data. [Supplementary materials are available for this article. Go to the publisher's online edition of Journal of Personality Assessment for the following free supplemental resources: the data set used to illustrate Poisson regression in this article, which is available in three formats-a text file, an SPSS database, or a SAS database.].

A Daily Diary Study of Joint Quit Attempts by Dual-Smoker Couples: The Role of Received and Provided Social Support

“…33 Moreover, zero-inflated negative binomial regression is used to model count data with an excess of zero counts. 34 The effect sizes for these models are rate ratios A c c e p t e d M a n u s c r i p t (RR) and are interpreted as the percentage increase (values > 1) or decrease (values < 1) in daily smoking for a unit increase in the predictor. 36 All analyses were conducted in R version 3.2.2 with the glmmADMB package for fitting generalized mixed models.…”

“…Thus, data were analyzed using a generalized linear mixed model that specified a negative binomial distribution with a logarithmic link function and zero inflation with a constant zero-inflation value only (ZINB). [33][34][35] The negative binomial model includes a random component reflecting the uncertainty about true rates at which an event occurs for individual cases while accounting for overdispersion. 33 Moreover, zero-inflated negative binomial regression is used to model count data with an excess of zero counts.…”

Reduced neonatal lung function and wheezing illnesses during the first 5 years of life

“…Results are presented as odds ratios, regression coefficients, 95% confidence intervals, p-values and incidence rate ratios, indicating relative change in outcome rates [20]. Associations were considered statistically significant if p,0.05.…”

“…Median values for the lung function parameters were compared between children with different phenotypes by Kruskal-Wallis test. The number of consultations was used as a count type outcome, best fitting a Poisson distribution [20]. Poisson regression was used to assess the association between Crs and Rrs and the number of primary care consultations for wheezing illnesses in the first 3 years of life and in the fourth and fifth years of life.…”

Perspective taking as a means to overcome motivational barriers in negotiations: When putting oneself into the opponent's shoes helps to walk toward agreements.

“…Thus, the subsequent analyses used the negative binomial regression procedure (Atkins & Gallop, 2007;Elhai et al, 2008;Gardner et al, 1995), which can be used to analyze count data and also accounts for overdispersion (i.e., variance is larger than the mean). We followed the procedure suggested by Coxe, West, and Aiken (2009) to test the effects of our independent variables (i.e., social motivation and perspective mindset in a first step) and their interaction (second step) on the counts of partial impasses.…”