PMC:5423899 / 11184-13220
Annnotations
{"target":"https://pubannotation.org/docs/sourcedb/PMC/sourceid/5423899","sourcedb":"PMC","sourceid":"5423899","source_url":"https://www.ncbi.nlm.nih.gov/pmc/5423899","text":"Statistical Methods\nThe weekly episodes of vomiting and diarrhea were analyzed with a mixed-effects negative binomial generalized linear model. Fixed effects in the model were week, diet, and the interaction between. The random effects were animal nested within clinic. For vomiting, the change from week 0 to weeks 1–4 for each diet and between diets was tested. For diarrhea, the change from week 1 to weeks 2–4 for each diet and between diets was tested.\nThe negative binomial models were also used to test for an effect of positive fecal testing for FCoV, C. perfringens enterotoxin gene, and FCoV antibody on vomiting and diarrhea occurrences by changing the fixed effects to week and positive/negative and retaining the same contrasts. This same model was altered to have a response of the total vomiting occurrences over the 4 weeks and fixed effects of diet and age. Vomiting was also modeled against the initial Spec fPL values using a negative binomial model, as before, however with the change in vomiting from baseline as the response and the fixed effect of Spec fPL results. The fecal scores were modeled using a linear mixed-effects model with fixed effects of day, diet, and the interaction between and a random effect of cat.\nThe binary data calculated from the vomiting occurrences was analyzed with a mixed-effects binomial model, using a logit link function. As before, the fixed effects were week, diet, and interaction between. Random effects were animal nested within clinic. The same hypothesis tests were also performed comparing the change between weeks within and between diets.\nAll contrasts performed using the mixed-effects models were corrected for the multiple comparisons using stepwise correction, to achieve a family wise error rate of 5% within each measure. The hypothesis tests of correlations were corrected using the false discovery rate method of Benjamini–Hochberg, to the 5% level. These analyses were performed in R v3.2.0 (10) statistical software with libraries lme4 (11) and multcomp (12).","divisions":[{"label":"Title","span":{"begin":0,"end":19}}],"tracks":[{"project":"2_test","denotations":[{"id":"28540291-18481363-43169047","span":{"begin":2032,"end":2034},"obj":"18481363"}],"attributes":[{"subj":"28540291-18481363-43169047","pred":"source","obj":"2_test"}]}],"config":{"attribute types":[{"pred":"source","value type":"selection","values":[{"id":"2_test","color":"#93e0ec","default":true}]}]}}