Modeling Associations Between Home Ownership and Mentally Unhealthy Days

• Activity Number: 584
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Hong Zhou*+
• Companies: Columbus Technologies and Services, Inc.
• Keywords: model ; Poisson ; negative binomial ; zero inflated ; home ownership ; mental health

Abstract

The association between home ownership and mentally unhealthy days (MUDs) was examined from the 2009 Behavioral Risk Factor Surveillance System (65,930 adults in 12 states). A significant percentage of respondents reported 0 MUDs in the past 30 days, suggesting excess zeroes. In addition, the variance was larger than the mean, indicating over-dispersion. We fit 5 different regression models: linear, Poisson (P), negative binomial (NB), zero inflated Poisson (ZIP), and zero inflated negative binomial (ZINB). Observed vs predicted probability plots suggested better fits of NB and ZINB models. After adjusting for 8 socio-demographic factors, home ownership was not significantly related to MUDs for the linear and P models. For the NB model, home owners reported fewer MUDS than non-home owners (ß=-0.16, p=0.03). Using ZIP and ZINB models, ß's for the logit component were significant [ß=0.19 (p=0.01) and ß=0.21(p=0.01), respectively], indicating that home owners were more likely to report 0 MUDs; the ß's were not significant for the NB/P count component of the models -p=0.92 and p=0.52, respectively. These analyses demonstrate the impact of model selection on the inferences.