Statistical comparison of dog and cat guard hairs using numerical morphology

Abstract 

Numerical features obtained from the guard hairs of dogs and cats (total 300 hairs per dog or cat) were statistically compared, in an attempt to discriminate between them. Using hairs from each of five mongrel dogs and cats, eight measurements (length (Len), maximum width (MaxWid), cross sectional maximum diameter, cross sectional minimum diameter, cuticular thickness of the cross section and three scale counts per 100μm length (observed at three positions: distal third (disSC), middle (midSC) and the proximal third (proSC) portions) and five indexes (hair width index (HWI), medulla index (MI), hair index, cuticle index and the difference in scale counts between the distal and proximal parts (defSC)) were examined. The range for each numerical feature overlapped each other extensively, and none of the features permitted a discrimination between dog and cat hairs, based on the values obtained. However, 12 numerical features, except for the midSC, showed a statistically significant difference between dog and cat hairs, as evidenced by a t-test. For the purpose of comprehensively comparing numerical features and statistically discriminating between dog and cat hairs, a discriminant analysis between the two were carried out using a multiple regression analysis. Four types of discriminant functions produced by combining over five numerical features were examined. Dog and cat hairs could clearly be discriminated using any of the discriminant functions. Species discrimination using the discriminant function permitted the species of a dog or cat to be determined, based on the overall morphologies of various numerical features. When experimentally collected test samples were investigated using the discriminant function using Combination-2, consisting of eight numerical features (Len, MaxWid, MI, HWI, disSC, midSC, proSC and defSC), all 10 cat hairs were correctly determined to be cat hair and 22 of 23 dog hairs were correctly identified. This discriminant function produced good results for species discrimination between dog and cat hairs.

Keywords: Dog hairs, Cat hairs, Numerically morphological comparison, Discriminant analysis