The Use of Informative Priors in Bayesian Modeling Age-at-death; a Quick Look at Chronological and Biological Age Changes in the Sacroiliac Joint in American Males.

AIMS Public Health, 4 (3): 278-288 DOI: 10.3934/publichealth.2017.3.278 Received date 28 March 2017 Accepted date 05 June 2017 Published date 07 June 2017 http://www.aimspress.com/journal/aimsph Brief report The Use of Informative Priors in Bayesian Modeling Age-at-death; a Quick Look at Chronological and Biological Age Changes in the Sacroiliac Joint in American Males Kanya Godde 1,2,* 1 2 Sociology/Anthropology Department, University of La Verne, La Verne, CA 91750, USA Department of Anthropology, University of Tennessee, Knoxville, Knoxville, TN, USA * Correspondence: Email: ; Tel: +1-909-448-4476. Abstract: The aim of this study is to examine how well different informative priors model age-at-death in Bayesian statistics, which will shed light on how the skeleton ages, particularly at the sacroiliac joint. Data from four samples were compared for their performance as informative priors for auricular surface age-at-death estimation: (1) American population from US Census data; (2) county data from the US Census data; (3) a local cemetery; and (4) a skeletal collection. The skeletal collection and cemetery are located within the county that was sampled. A Gompertz model was applied to compare survivorship across the four samples. Transition analysis parameters, coupled with the generated Gompertz parameters, were input into Bayes’ theorem to generate highest posterior density ranges from posterior density functions. Transition analysis describes the age at which an individual transitions from one age phase to another. The result is age ranges that should describe the chronological age of 90% of the individuals who fall in a particular phase. Cumulative binomial tests indicate the method performed lower than 90% at capturing chronological age as assigned to a biological phase, despite wide age ranges at older ages. The samples performed similarly overall, despite small differences in survivorship. Collectively, these results show that as we age, the senescence pattern becomes more variable. More 279 local samples performed better at describing the aging process than more general samples, which implies practitioners need to consider sample selection when using the literature to diagnose and work with patients with sacroiliac joint pain. Keywords: Lovejoy method; auricular surface; transition analysis; pelvis; survivorship Abbreviations: PDF—Posterior Density Function; HPDR—Highest Posterior Density Regions 1. Introduction It has been established that skeletal age indicators alone do not adequately describe and capture chronological age [1–4], although they may be decent indicators of biological age. In order to contend with the differences in age types, the application of Bayesian statistics, which takes into account a prior age-at-death distribution, can help to model biological age so that it captures chronological age well (e.g., 1–7). One of the first considerations for trying to ascertain chronological age from the skeleton with statistical analysis is identifying an appropriate prior sample from which to draw an age distribution in Bayesian analysis. Populations age differently skeletally [1,3,7–9], which necessitates finding a prior sample with a relatively similar age distribution, usually from a source population comparable to the target population (the population to which a practitioner seeks to assign chronological ages). The informative prior can also be derived from the target population, wherein a portion of the population from which chronological age is to be calculated can be used as the informative prior, and is referred to as the forensic approach in the literature (c.f., 2). The knowledge of this prior age distribution in a Bayesian analysis strengthens the accuracy of age estimations assigned to different phases of a skeletal indicator, although past work has established that there is some leeway in the fit of the prior sample to the target sample from which age is to be estimated [1]. Biological indicators of age are important to understand for historic and forensic populations. Knowledge about how these indicators change as we age can inform practitioners of joint changes that affect older adults, which in turn apprises application of age-at-death estimation methods. In particular, the sacroiliac joint is a source of pain for adults as they age due to the development of ankylosing spondylitis and changes to the cartilage that cushions the joint. As one ages, the sacroiliac joint transitions to become rougher and more irregular [10,11]. Changes to the cartilage is asymmetric; degeneration of the iliac surface occurs at an earlier age than the sacral surface [11]. Movement at the joint can become restricted when a person is in their 50s due to cartilaginous changes that result in an increase of collagen and fibrous ankylosing [11]. These senescent alterations continue into a person’s 70s [11]. The morphological changes at the sacroiliac joint are the features that are captured in AIMS Public Health Volume 4, Issue 3, 278-288. 280 biological indicators of age; each phase describes a set of characteristics that we transition through as we age. Two popular methods of sacroiliac, or auricular surface, skeletal aging include an 8-phase [10] system and a composite score method [12]. In this research, I utilize the Lovejoy et al. [10] 8-phase method as a measure of the applicability of prior samples. It is of interest in this research to look at different representative levels of prior samples to ascertain which is the most appropriate for selection in Bayesian analyses of age-at-death. The potential pools of age-at-death structure data can be found in skeletal collections, cemeteries, county records, and country records, to name a few. Here, I examine the age-at-death demographics from an American skeletal collection developed mostly from willing, local donors, a cemetery local to the skeletal collection, the county in which the collection and cemetery reside, and the country in which the collection, cemetery, and, county are located. The auricular surface indicators were recorded from individuals in the skeletal collection. Thus, the forensic approach, wherein a subset of the target sample is used as an informative prior, will be called upon in this analysis. This paper takes a biodemographic approach to investigating age and age-related changes in the skeleton. While this paper focuses on skeletal indicators of senescence, these data inform us of the aging processes humans undergo, which provides insight into longevity. The age-at-death distribution of several samples are examined and tested for differences that might impact age-at-death information derived from the skeleton, which impacts our understanding of biological and chronological age. Based on the literature, I hypothesize there will be small differences in the age-at-death distributions of the four samples. The performance of each sample will be measured by the calculated accura (...truncated)