| dc.description.abstract | Data was gathered during an in vivo experiment concerning periparturient supplementation with palmitic and oleic acid and their effects on blood parameters and reproductive performance in heifers. This resulted in a dataset with repeated measures per cow within two experimental groups. Mixed models were used for the analysis of the dataset since they are specifically designed to deal with this structure within the data. Even then, choosing the right distribution that matches the data is necessary to properly analyse it, as seen in the haptoglobin model. This was solved by using a t-distribution with 2 degrees of freedom instead of the normal distribution that fitted the other blood parameters. For the reproductive parameters, a Poisson distribution was used for the number of aspirated follicles and the number of collected oocytes while a binomial distribution was used for the cleavage and blastocyst rates. To determine which independent variables should remain in each model, a step-down technique was used combined with Akaike’s information criterion. The outcome of this processes varied between all independent variables, and an interaction term for the follicle count and no independent variables for the cleavage rate. With the small sample size for this experiment (eleven cows in total) and some missing values in the dataset, some caution is necessary when interpreting the results. For reporting the results, the use of confidence over p-values can provide more valuable information over the distribution and variability of the data, as well as a better indication of the biological or clinical relevance. | |
