Factor screening in a 12 Run Plackett-Burman design assuming four active factors

Abstract

In this thesis we perform factor screening in a non-regular two-level design by reducing the number of possible sets of active factors to a certain number. The 12 Run Plackett-Burman(PB) design with four active factors is mainly concerned. Our proposed method works through picking up the 6 effects with the highest absolute value out of 10 in each projection model. To evaluate this method, we used the same example as was used in Tyssedal and Shahrukh\cite{tyssedal2016factor} where variable selection methods such as $AIC$, $F$ test and $\bigtriangleup R^2$-method used on projection models. A real example is included at the end to show how our proposed factor screening method can be done in practice.