Interaction Analysis with Covariate of Noise Factor  

Oleh:

Goto, Kensuke ; Nagata, Yasushi

Jenis: Article from Proceeding
Dalam koleksi: 12th ANQ Congress in Singapore, 5-8 Agustus 2014, page 1-7.
Topik: Robust parameter design; Interaction Analysis; Linear regression model; Dummy variable; Monte Carlo simulation
Fulltext: QP2-2.4-P0254.pdf (330.63KB)

Isi artikel

Robust parameter design aims to make the design insensitive to noise factors by choosing optimum levels of controllable factors. This design is conducted by finding the interaction of noise factors and control factors. In general, it is necessary to control levels of noise factors. However, there exist noise factors which cannot be controlled as levels in experiments. On the other hand, there exist noise factors which can be observed as covariates. Hirano and Miyakawa(2007) proposed a method based on the linear regression model to analyze the interaction of noise factors and control factors when noise factors are observed as covariates. Koyano(2013) proposed a method to analyze by using propensity scores based on noise factors when multiple noise factors are observed as covariates. However, since these methods analyze by combining noise factors, they cannot detect which noise factor has influence on control factors. Furthermore, if some noise factors which have no effect are included, influence of existing interaction effects is weakened. Therefore, it might not be detected accurately. This paper discusses interaction analysis of noise factors and control factors when multiple noise factors are observed as covariates. We propose an improved method that analyzes by distinguishing noise factors. Furthermore, we investigate estimate accuracies by Monte Carlo simulations under several data models. It is shown that the improved method is able to detect the interaction of an individual noise factor. By contrast, the preceding methods are not able to detect it

Opini Anda

Klik untuk menuliskan opini Anda tentang koleksi ini!

Kembali