Selecting the Number of Independent Variables Factors

Having known dispersions of the independent variables of, the contribution of each of them to the total resulting dispersion can be appraised as follows  

Then the test ofQcer can make a decision about taking into account or rejecting the appropriate factor, after calculation of the values of the relative contribution of the variable into the total dispersion D  

The decision-making process described above is acceptable only to the linear tasks. However, considering zirconia gas sensor testing, where the volume of main testing is connected to appraisal of the sensor metrological characteristics, the quasi-linear tasks usually take place. Therefore, despite the restriction by linearity as well as a certain level of subjectivism, this way is relatively easy and clearly excludes variables, the insufQcient value of which at comparison Yj is unquestionable. 

The distribution laws of values of the influential factors are accepted to be uniform at calculation of of. The value of the total dispersion D is equal to 0.1433. The multicriteria active experiment has to be done for more accurate determination of valuable dominating factors. The results of this experiment are used in dispersion analysis. Then the residual dispersion, stipulated by the error of experiment, is calculated as follows  

taking into account the ratio Di/Dj s, criterion of statistical significance of the Fisher dispersion can be applied for further calculation. In (7.3), Dq is the total dispersion of the results relative to the total average value, Dj is the dispersion of the average results value by all r levels of the first factor, and D is the dispersion of the average results value by all r levels of the factor. More detailed explanation of the dispersion analysis theory can be found in [6, 7]. 

---