| The scale and intensive process of pig production have promoted the importance of production data management.The existing production management system can achieve simple performance data rendering and comparison analysis,but lack of analysis of the factors related to the causes of the results..This study explored the method of establishing for large-scale data of large-scale pig production,and used to analyze the reproductive performance data of 16 farms for 5 years.Modern scale pig farms usually make PSY as an important index to measure the integrated production level.Litter size,piglet birth weight and pre-weaning mortality are the key to influence the PSY.Different farms belonging to the same company will present different performances,have obvious field-level management characteristics.This study is based on these key reproductive performance,to a large breeding group of 16 farms during 5 years,and explore and establish the appropriate multi-level model analysis method,analyses the five years of production data,and in the establishment of the model and analysis of some statistical problems encountered in the comparative study,found that some of the conclusions have guiding significance to practical production,and tries to set up production index prediction model for the forecast of the future.This study consists of three parts.At the first part,a two level hierarchical model based on the sow farrowing nested in farm level was established,and the factors affecting the litter size and the birth weight of piglets were analyzed.At the second part,a two level hierarchical model based on the weaning batches nested in farm level was established,and the factors affecting the preweaning mortality and weaning weight were analyzed.At the third part,production index prediction model was established,and the easy-to-use operation platform in the actual production was researched.The main results were as follows:1.A multi-level model was established to evaluate the factors influencing the sow litter size and the birth weight of piglets.1)For the data with hierarchical characteristics,the multilevel linear model should be choosed instead of the general linear model.For litter size and such counter variables,when the sample size is big enough,the analysis results that use multilevel linear model and multilevel poisson regression model had little difference.Effects of sow feeds supplemented with OEOs on litter size and piglet birth weight,under the multi-layer linear model and general linear model,the analysis results have obvious difference.2)The extent which the variables account for the results indicates that there are still important factors that are not included in the model,such as disease.It is suggested that information on the occurrence of diseases should be included and quantified in production management data input.3)Animal factors such as sow breed,parity have significant influence on litter size and newborn weight of piglets.Sow breed had significant effect on total litter size,live born,healthy born and body weight born(P < 0.01).The litter size and birth weight had no difference between the binary and Large White.The total litter size from high to low in sequence of binary cross-breeding/ Large White,Landrace,Duroc.The Duroc had the total litter size lower 3.29,healthy litter lower 2.15,birth weight lower 0.12 compared with the Large White.The parity also had significant effect on total litter size,live born,healthy born and body weight born(P < 0.01).4)Sow feeds supplemented with OEOs did not affect the total litter size and the born alive,but increased the number of born healthy by 0.52(P <0.05),at the same time to improve the piglet birth weight by 0.0178(P < 0.05).The essential oil treatment had obvious storage effect on the reproduction performance of sows(litter size and birth weight of piglets).5)The management level significantly affectted the total litter size and born alive,but had very significant effects on born healthy.Compared with the good management,the poor management had lower born healthy by 1.56 and less birth weight by 0.06.The sow source field has significant effects on total litter size,but for the number of born alive and the number of born healthy had no effect.6)In June-September,number of born healthy and birth weight decreased compared with February-May and October-January.2.A multi-level model was established to evaluate the factors influencing the the mortality and weaning weight of piglets.1)The management level,the sow feeds added OEOs,the year and stage of piglet birth significantly impacted on mortality.Compared with good management,poor management significantly increased the mortality(SQRM increased by 1.62)and reduced the weaning weight by 0.74.Sow feeds added OEOs reduced the pre-weaning mortality(SQRM reduced by 0.65)and improved the piglet weaning weight by 0.49.The lowest preweaning mortality appeared on the piglets born in the hot season in June-September and highest one appeared on the piglets born in the cold season in October-January.The highest piglet weaning weight appeared on the piglets born in February-May with relatively suitable temperature.2)The model fitting comparison showed that the multi-level linear model is better than a single linear model.The conclusion may not accurate when using the general linear regression model analysis multi-level data.By introducing a variable to reduce the proportion of the total variance,suggested that the main contribution of sow feed added OEOs is to improve the piglet growth rate,and the main contribution of good management is to reduce the piglet mortality.3)The SQRM and weaning weight of piglets born in three different phases had some degree of heteroscedasticity.In February-May,mortality and weaning weight had the largest fluctuations.It seems easy to ignore the follow up of management measures in temperature appropriate season.3.The realization integration of the production index predictionAccording to the previous assessment of the effect values of various factors,based on multivariate linear regression forecast model was set up.According to the change of the factors,the future value of the production indexes was predicted.Finally,the prediction effect has carried on the inspection and evaluation.Multiple linear regression forecast model was established in this study,its forecast effect is very good,can be used as a production index prediction is an important auxiliary means for production.The prediction model software was compiled about the related indicators(litter performance of sows and piglets lactation piglet mortality and weaning weight).Multifarious operation are integrated into the software of statistical process module,the program built-in,modular,interface of operations and making a fool of,meet the requirements of the production personnel for a quick and play. |
PDF Full Text Request