Discriminant analysis is a useful analytic method in maximal seperation of group means, finding more meaningful factors on group discrimination and group seperation by discriminant function. This analysis can be applied to multiple
electrodiagnostic
parameters for diabetic and nondiabetic subjects to reveal the essential dimension along which key neuropathic differences occur between these groups.
In this study electrodiagnostic data of 121 diabetics and 30 controls was obtained and 25 electrodiagnostic parameters including distal latency, amplitude, velocity, F-wave minimal latency and H-latency were used in a stepwise discriminant
analysis. At
first, discriminant analysis between control and diabetic group was used to reveal a few meaningful parameters and produce a discriminant function seperating two groups. And then, discriminant analysis between control and non-neuropathic
diabetic(I)
group was performed to find parameters reflecting early change of polyneuropathy.
The discriminant analysis between controls and diabetics revealed 6 meaningful parameters(eigen value=0.5559, p=0.00) and analysis between control and non-neuropathic diabetic group(I) reveals 9 parameters(eigen value=0.7196, p=0.0003). The
H-latency
was the most significant parameter H-latency, median motor conduction velocity and ulnar F-latency are reproducible parameters in two analyses. The sensitivity and specificity of the discriminant function seperating the control and the diabetic
group
were 78.5% and 96.7%, respectively, and between the control and the non-neuropathic diabetic group(I) were 82.5% and 76.7%, respectively. Though the statistical complexity is recognized, we belive discriminant analysis is useful in the detection
of
diabetic neuropathy and is needed for classification of many diabetics into neuropathic and non-neuropathic categories with a few electrophysiological parameters.
|