New approach to the study of the pharmacological effect (pentobarbital sleeping time) in relation to individual properties of the organism (psychophysiological parameters of behaviour in rats) using multivariate statistical techniques
Plevová, J.; Formánek, J.; Holoubková, E.; Tikal, K.
Acta Universitatis Carolinae. Medica 24(1-2): 91-109
1978
ISSN/ISBN: 0001-7116 PMID: 753098 Document Number: 127206
The relationship between individual characteristics of the vertical and horizontal components of exploratory activity (symbols v and h) and individual sum totals of v + h (symbol o), frequency of defecation (symbol d), body weight (symbol w) and duration of pentobarbital narcosis, so called pentobarbital sleeping time (symbol t) in rats was studied to assess the possibility of predicting the variable t on the basis of the other variables (v, h, d, w). Factor analysis, standard correlation analysis (Pearson.sbd.Bravais), multiple regression and multiple correlation analysis permitted an insight into the interrelations between the 5 variables v, h, d, w and t, and also permitted the formation of some predictive hypotheses for the variable t. Coefficients of standard correlation (Person.sbd.Bravais) between the individual values of t and v, h, o, d or w were significantly different from 0 in the cases of t and v and t and o (P < 0.05). The grouping of animals into subgroups according to their individual combination of characteristics d and o also showed a statistically significant influence on the mean values of t (P < 0.01) in the subgroups. The use of the relevant technique view to the problem of prediction of t, i.e., the multivariate regression and correlation analysis, revealed that the partial regression and the multivariate regression of t on the other 4 variables v, h, d, w could not be proved; thus the coefficient of multivariate correlation was of no practical significance and a corresponding prediction of t on the basis of the known v, h, d, w was not possible. This material was used to demonstrate some general methodological principles which should be regularly applied in an analysis of the structure of individual properties of the organism, in the formation of predictive hypotheses and their verification. Detection of non-linearities and of less common types of regression, standardization of data, and the inappropriateness of selecting extreme groups of individuals were emphasized.