[Linear discriminant analysis in the diagnosis of metabolic diseases in cows].

Studied were a total of 480 high-producing cows to determine the signs of highest information value by which the animals could be divided into groups of normal ones and such of animals presenting various metabolic disturbances: osteomalacia, ketosis, and liver dystrophia. Used were as many as 25 qualitative signs as established through the anamnesis and the clinical and paraclinical investigations. It was found that by means of 8 of these signs only (age, erythrocytes, blood sugar, vitamin A, carotene, total protein, total bilirubin in the blood serum, and pH of the urine), employing a linear discriminant function formed with them one could properly diagnose the disease condition of 83 per cent of all investigated animals. It was also established that the increase on the number of signs to form the linear discriminant function did not lead to an essential change in the preciseness of diagnosis.