The Algorithm of Using the Logistic Model for Binary Dataanalizis in Ecology Investigations

Aim. The GLM (Generalized Linear Model) algorithm has been created for working with binary data on all analysis steps, which would be possible for all users without special IT skills. Methodology. We used wide distributed algorithm statistically analysis which includes three main steps: biology hypnotizes formulation and data collecting, investigation of data, fitting and checking finally models. As data example were choose forest taxation features of Oak petiolate (Quercus robur L.) from 15 ecoregions. The average trunk diameter has been simulated according to age, high, trunk number and geographically location of sample area. Because features of dependent variable didn’t allow choosing classical regression analysis we simulated of probability the excess of 30-cm threshold of trunk diameter with logistic regression using. For quality assessment of models we used different deviances, i.e. difference between depended variable and predicted values (residuals), calculated specific quantile residuals for GLM, nonparametric tests for nested models and model parameters potentiating. Results. There was showed the dependence trunk diameter on age and trunk number, as well as latitude of sample area. The existing algorithm has been supplemented by power test and assessment of prediction independent variable by kappa-coefficient and ROC-curves. The new algorithm allows simulating variables which don’t fit to demands of lineal regression, to comparison models, to assessment the quality of prognosis and a volume of minimal sample. Research implications. This work has a methodic direction. There has been showed for logistic models creation there are enough functions from RStudio core excluding the minimal sample assessment.