Investigations of the effect of v-rays on radiation damage in materials during irradiation in a reactor [1][2][3] serve as a source of illustrative proving material, but they do not make it possible to draw quantitative conclusions. Specifically, they do not answer completely the question of which independent variables are statistically significant and how strongly they influence the properties of the irradiated material. For this, a statistical analysis of the results of irradiation of diamond in the working channel of a materials reactor was performed.The experiment was performed under conditions of relative scales for the reactor radiation flux density, since irradiation was performed simultaneously in one channel of the reactor. The fast-neutron flux density ~o n was measured with the aid of threshold detectors, the ~,-ray flux density ,p.~ was measured by measuring the radiation heat release in the constructional material by the statistical calorimeter method with a correction for the effect from fast neutrons. The irradiation temperature Tit r was determined according to the kink in the isochronal annealing curves for diamond [4]. The experimental results are presented in Table 1.The expansion of the diamond crystal lattice under irradiation depends on certain factors whose complete subset is still unknown. Only certain of the presumed factors or their functions are known. To solve the question of the statistical significance of these variables, we employed the method of stepwise analysis of multidimensional linear regressions [5]. This procedure makes it possible to determine the subset of variables corresponding to the best regression.The analysis is based on expansion of the unknown function near the chosen point in a multidimensional space of independent variables in a multiple power series whose coefficients are found by the least-squares method. The individual terms of the model adopted are included in the equation successively in the order of decreasing numerical values of the squared partial correlation coefficient between the variables and the response, which are determined prior to each step. The successive variables are included until the rule for stopping the procedure is triggered, i.e. as long as Fafte r > Ftabl e (1, np --I, 1 --or), where F is Fisher's criterion, n is the sample size, p is the number of variables introduced into the equation, and a is the degree of significance. After each step the null hypothesis H0: fir = 0, r < p is checked for all previously introduced variables, except for the last one, by comparing the corresponding partial F criteria with their tabulated values for ot = 0.05. Many such programs have now been published [6]. We used one of these programs in a modified form.On the basis of a preliminary analysis and general considerations it was concluded that not only the independent variables from the table but also some of their functions should be introduced into the model as regressors [5]. Transforming the variables, centering, and introducing new notation, we obta...