Purpose: Medical experiments frequently involve observations of many variables. For instance, the results of a blood chemistry test are a set of such multivariate data. The study attempts to diagnose the occurrence of scoliosis based on the results of biochemical analyses of blood serum samples.
Methods: The following hormones were measured: FSH, LH, E2, PROG, HGH and PTH. Osteocalcin, calcium, phosphorus and vitamin D were also measured. Discriminant analysis and decision trees were applied to determine which of the measured parameters would allow the studied set to be divided naturally into four groups of patients.
Result: After such a division, these techniques allowed for prognoses, and thus, for assigning new data to the four classes. The results of advanced statistical analyses are presented as classification functions and in a graphical form as a decision tree, as well as a set of decision rules. The study showed that diagnosing scoliosis is possible based on five (LH, E2, PROG, calcium and osteocalcin) out of ten available results of biochemical measurements.
Conclusions: Statistical analysis allowed for classifying new medical cases with high probability, which may even increase after new data are introduced into the dataset and corrected classification systems are generated.