A STUDY OF THE EFFECT OF INPUTS ON LEVEL OF PRODUCTION OF DAIRY FARMSIN QUEENSLAND - A COMPARATIVE-ANALYSIS OF SURVEY DATA

Multiple linear regression models able to estimate total farm milk pro duction from nutritional inputs were developed from farm survey data p rovided by dairy farmers in Queensland, Australia. These models were s pecifically developed for inclusion in a decision support system that could provide dairy farmers with an annual milk production estimate, t hus enabling them to compare their production with an average farm usi ng the same inputs in their region. Separate models were developed for each of I regions in Queensland and an additional model was developed for farms producing greater than 750 kL of milk per farm per year. Th e models were tested on dairy farms in Queensland by using the decisio n support system on farms that were not involved with initial model de velopment. The partial regression coefficients for the models were bio logically sensible and, apart from some minor interactions between ind ependent variables in 2 regions, were additive. These interactions wer e not included in the final model in the interests of parsimony, ease of explanation and a need to provide transparent models within the dec ision support system. The coefficients of determination (R-2) for the models varied from 79.9 to 88.3%. Forward-feed artificial neural netwo rk models were also used to confirm the relative accuracy of the multi ple linear regression models and to allow for any interactions or non- linear functions in the data and to show that the simple equations are more appropriate for a farmer-orientated decision support system.