Semester of Graduation
Veterinary Diagnostic and Production Animal Medicine
First Major Professor
Master of Science (MS)
Veterinary Preventive Medicine
Figuring out what probability distribution your data fits is critical for data analysis as statistical assumptions must be met for specific tests used to compare data. However, most studies with swine populations seldom report information about the distribution of the data. In most cases, sow farm production data are treated as having a normal distribution even when they are not. We conducted this study to describe the most common probability distributions in sow herd production data to help provide guidance for future data analysis. In this study, weekly production data from January 2017 to June 2019 were included involving 47 different sow farms. We evaluated 14 variables and report descriptive statistics, including mean, median, standard deviation, range, minimum, and maximum. Variables were also analyzed to identify goodness of fit for selected distributions. Goodness of fit was identified based on the p-value of previously validated test results. A total of 15 distributions and 2 transformations were tested. Any p-value larger than the pre-set level (0.05) was accepted which means the variable fits the candidate probability distribution. The result demonstrated that when data from all farms were combined, for variables farrowing rate (p-value=0.110) and average stillborn (p-value=0.179), the Johnson Transformation was the best approach to then fit a normal distribution. All the other variables did not fit any of the distributions or transformations tested. Then each farm’s data was analyzed individually. Results showed that herd-level variables (i.e., not aggregated) often fit to multiple distributions. Very few variables were considered as normally distributed. However, most variables were fit to normal distribution after being transformed using the Johnson or Box-Cox transformation methods.
Embargo Period (admin only)
Tong, Hao, "Identifying Probability Distributions of Key Variables in Sow Herds" (2020). Creative Components. 617.