Στατιστικά κριτήρια για την αξιολόγηση απλών (regular) και σύνθετων (nonregular) παραγοντικών σχεδιασμών με δύο επίπεδα
Statistical criteria for the evaluation of two level regular and nonregular factorial designs
View/ Open
Subject
Factorial experiment designs ; Στατιστική ; Γραμμική ανάλυσηAbstract
Fractional factorial designs with factors at two levels are the most commonly used factorial designs in various experimental fields. The economy of smaller run size compared with factorial designs at more than two levels makes fractional factorial designs with factors at two levels attractive for studying a large number of factors. As part of orthogonal factorial designs, the fractional factorial designs can be generally classified into the regular fractional factorial designs, that have simple aliasing structure in which any two effects are either orthogonal or fully aliased and the non¬regular fractional factorial designs, that have complex aliasing structure in which effects are neither orthogonal nor fully aliased. Furthermore, the criteria for selecting optimal regular or non-regular fractional factorial designs can be classified into two categories: the design-based and the model-based criteria. It is of practical use and scientific interest in statistical process control to rank and compare both regular and non-regular factorial designs in a systematic manner. The main purpose of current thesis is the presentation, the detailed description and the application of some presented in literature criteria used for the evaluation and selection of the optimal regular and non-regular fractional factorial designs with factors at two levels. A complete evaluation of the full lists of nonisomorphic orthogonal arrays with 28 runs and 5 two-level factors, with 32 runs and 4 two-level factors, and finally, with 32 runs and 5 two-level factors is presented.