Short communication: statistical models for the analysis of coagulation traits using coagulating and noncoagulating milk information.

Aims of this study were to propose statistical models for the analysis of rennet coagulation time (RCT) suitable for making use of coagulating and noncoagulating (NC) milk information, to estimate heritabilities and to obtain rank correlations for sire merit. A total of 1,025 Holstein cows (progeny of 54 sires) reared in 34 herds were milk-sampled once. Data were analyzed using 4 alternative models: a standard linear (SLM), a right-censored linear Gaussian (CLM), a survival (SUM), and a threshold (THM) model. Model SLM analyzed coagulated milk records only, whereas analysis with CLM or SUM considered information of NC samples as censored records. Model THM analyzed occurrence of milk coagulation as a dichotomous trait. An artificial censoring scenario with an endpoint at 18 min (SET18) was considered after the rearrangement of the timeframe originally used for the observation of RCT (SET31). Heritabilities ranged from 0.12 to 0.25. Correlations of sire rankings ranged from 0.23 to 0.92. Differences in sire rankings between SLM and CLM or SUM increased when the proportion of NC records increased. Correlations between sire rankings obtained for SET31 and SET18 were high for CLM and SUM, indicating that rankings provided by these models tended to be stable even when a large fraction of samples with observed RCT was re-classified as NC milk. Results indicate that CLM and SUM are more suitable than SLM and THM for the analysis of coagulation ability when data contain NC milk information.