Date of Award
Doctor of Philosophy
Ralph G. O'Brien
The purpose of this research is to develop a procedure to initiate process control. Traditional methods of process control do not allow testing until an initial group of samples (usually 20 or more) is drawn. In addition, the testing for control worth respect to the process average is dependent on evidence that the variability in the process is stable.
The procedures developed and studied here allow immediate testing on the process average without assuming that the process is in control with respect to variability. Through the use of a sequential testing procedure, data from a current sample is tested to determine if it indicates that the process mean is updated.
Two sets of test statistics are considered for the sequential procedure. Both are modifications of the traditional test statistic used to test the equality of the means of two independent samples. One set of test statistics modifies the estimate of σ2 by pooling the sample variances; the other set of test statistics uses a version of the Welch t with estimated degrees of freedom.
An empirical study is conducted to evaluate the proposed procedure using all versions of the test statistics. Comparisons are made between the proposed procedure and current practices. Finally, extension of this procedure to estimation of other parameters and to applications outside statistical process control are suggested.
Melton, Kim Ingrid, "A Procedure for Initiating Process Control. " PhD diss., University of Tennessee, 1986.