JMP platforms and statistical methods

Distribution:
Histograms, box plots, summary statistics

Graph Builder:
Scatter plots, comparative dot plots, linear regression fit, confidence intervals, bar chart

Formula Editor:
Creating new variables

Multivariate:
Scatter plot matrix, correlations

Fit Y by X:
Simple linear regression, one-factor ANOVA

Fit Model:
Multivariable linear regression

Objective

Build a multiple linear regression model to identify the factors that impact the amount of unwanted particles in a semiconductor cleanroom and model the relationships choosing how to handle correlation among the predictor variables.

Problem statement

Semiconductor manufacturing requires very clean environments to help prevent small particles in the air from contaminating the wafers. Though all the workers use clean suits and follow certain procedures before entering a clean room, the presence and activity of people in those rooms are known to be a major source of these unwanted particles.

To better understand how human activity in a clean room affects the amount of particles present, data has been collected for the past 45 days measuring the amount of particles detected in a clean room under a variety of conditions. The study was conducted during typical operations so it was not possible to deliberately choose the number of technicians that would be entering/exiting the lab and how frequently, as well as how much activity (amount of movement technicians engage in) when in the lab. Two experimental factors were chosen for the study (type of clean suit and a pre-entrance cleaning procedure). These two factors were able to be deliberately changed. Each day, the technicians were instructed to wear one of the three different types of clean suits and to perform or not perform a particular pre-entrance cleaning process, which included face washing and hair brushing, among other steps to shed potential particles.

A set of measurement instrumentation was installed throughout the room positioned at two heights (ground and head level). The instruments take air samples in multiple locations throughout the room across various points in the day. The total amount of particles counted in aggregate throughout the day will be used as the response data.