Influence of prior distributions and random effects on count regression models: implications for estimating standing dead tree abundance

摘要: The presence and abundance of standing dead trees (SDTs) in forests is typically characterized by an excess number zeros high variation. variability inherent SDT data naturally leads to the assessment novel quantitative methods represent populations their role contributing forest ecosystem structure. This analysis assessed performance count regression fit with Bayesian mixed-effects models that estimate SDTs (all a diameter at breast height \({\ge }12.7\hbox { cm}\) found on plots 0.07-ha size) over 17,000 inventory across US Lake States (Michigan, Minnesota, Wisconsin). Random effects used independent variables basal area, mean annual temperature, plot ownership (i.e., publicly- or privately-owned) as fixed type random effect outperformed method which fixed-effects, alone. Standard zero-inflated negative binomial were effective accounting for overdispersion present (variance/mean ratio counts was 72.6), whereas Poisson not. calibrated new population SDTs. Employing informative prior distributions from developed model led improved estimates reducing root square error seven percent. highlights approach uses existing representing averages while calibrating local arrive region.