Page 28 - 6686

4.3 Analysis of equation of regression. Evaluation of
variance of reproduction or error of experiments

After determining coefficients of equation of regression,
begins statistical analysis of the equation that consists of three
stages:
- Evaluation of variance of reproduction or assessment of
error of experiment;
- Evaluation of significance of regression coefficients;
- Evaluation of model’s accuracy.
Errors of an experiment or variance of
2
reproduction S о is measured for experiments, which are
conducted parallelly. Before calculating error of experiment,
first of all, make sure that variance of the results of experiments
does not exceed certain value. For this purpose, row
2
variance S i is calculated and checked for being
homogeneous. Row variance is determined by the formula:
2 1 m 2
S u    y uj  y u  . (4.10)
m  1 j  1

2
Homogeneity of variance S і is confirmed by Cochran’s
criterion, and calculated value is obtained by the formula:
2
S
G  u max ,
p N 2 (4.11)
 S u
u  1
2
S
where u max is maximum value of calculated row
variance;
N 2
 S u
u 1 is the sum of all variances for N rows of
matrix of planning.
Values of Cochran’s criterion at q = 0.05 are given in
28