Statistical Analysis

Two linear mixed effects models were constructed. The first model was designed to predict the temperature of stacked poultry litter piles. Data from continuous temperature probes that recorded a temperature every 10 minutes were used to construct these models. Data from Pile 5 is not included because we were forced to pull the instrumentation from that pile due to extreme heat. The first model uses temperature data averaged by hour. Fixed effects were time, height of the probe measurement from the base of the pile, depth of the probe measurement from the side of the pile and the pile number (1, 2, 3, 4, or 6). Random effects are repeated temperature measurements on sections within piles. Pile 1 is the referent pile in both models. Wald's t-test was used to assign probabilities to the various terms. The second model was design to predict which physical parameters, water activity, moisture, pH were driving the heating profiles and/or accounting for the differences between pile temperature profiles. Model 2 was also a linear mixed effects model. We used a 24 hour mean temperature (a daily mean for the continuous probe data) in this model. Fixed effects were day, height, depth, water activity x 100, pH, and the interaction terms Aw x pH and day x depth. Random effects were as for Model 1. The details of the models are given in the Appendix section.