Comparison Between Different Regression Techniques on Tablet Compression Machine Batch Record Data
Currently, pharmaceutical industry has experienced significant growth and tends to continuously grow in the future. Therefore, pharmaceutical companies need to increase their competitive abilities. Yield performance is one of the many factors that impact on improving pharmaceutical manufacturing performance in term of competitive capability and company profit increase. Loss of product during manufacturing process is the main factors that effect on yield performance and machine parameters may be one of many factors that effect on product loss. In this research, we aimed to evaluate the relationship between product loss and machine parameters of tableting process by analyzing first degree and second degree of data with none selection and three stepwise selection procedures of multiple regression. Significant factor, R-sq, and lack of fit (p-value) were observed during study. The result clearly showed that the interaction between tablet compression machine parameters and product loss was not linear. None the less, we were still able to extract significant factors when interpreting with none selection and backward selection procedures. The one factor that came out as significant from these two methods was the interaction among tableting machine speed (X1) and feeder (1-layer) speed (X2).
Keywords - multiple regression, tablet compression machine, pharmaceutical manufacturing, stepwise analysis.