Prior to the commencement of deregulation from 1 July 2000, the Australian Dairy Research and Development Corporation conducted a large-scale telephone survey of 1826 Australian dairy farms to examine the current on-farm management practices in relation to milk production and farm and farmer demographics. The questionnaire results from the 214 dairy farms in the sub-tropical region of South East Queensland and Northern New South Wales were analysed (Zamykal et al. 2007) to uncover those significant inputs that affect milk production.
In order to uncover management practices and the underlying but unobservable variables (random quantities) that significantly contribute to milk production, the data was analysed using two major techniques. Firstly, the number of cows a farm possesses is obviously shown to be a significant predictor of milk production. This strong overriding linear relationship is removed from the analysis by using the residuals from this regression as the new dependent response variable. Therefore, the residuals are the effect of herd size removed or simply the herd size effect (HSE). The original variables are then regressed against the residuals and the variables which significantly predict the residuals are then highlighted. Secondly, factor analysis was used to extract a reduced number of factors from the sample correlation matrix R, in the absence of the HSE. It was anticipated that the model derived from the factors would consist of a few interpretable factors that explained some underlying but unobservable random quantities hidden within the original variables.
These new factors where then regressed against the residuals derived from the initial regression with the intent of highlighting those significant unobservable random quantities. Both models produced from the analysis revealed a number of similarities and differences. Comparison of the two linear models reveals that age or experience is negatively associated with predicting milk production in the absence of the HSE. The use of irrigation was also found to be an important component in predicting the residuals. Comparison of other variables and components revealed differences in the composition and interpretability of both models. The factor model allowed the analysts to discover an unobservable random quantity that may influence the inclusion of standard variables in the initial regression model. The inclusion of such a factor allowed the analysts to compare the model derived from standard variables and asses whether both models described the same quantities. An important outcome of the analysis was to reveal and contrast the variables or quantities that significantly impact the manager’s ability to increase milk production in the absence of the traditional increase in herd size. This is useful for improving the efficiency of dairy farm operations within the targeted region.
- multiple regression,
- factor analysis,
- varimax rotation,
Available at: http://works.bepress.com/mike_steele/5/