Sunday, November 15, 2009

Best SPC control chart option?

Here's the scenario: Individual customer satisfaction survey scores (measure is % of surveys with 'excellent' score) collected more or less at RANDOM intervals throughout each quarter yet summarized and reported on a QUARTERLY basis as a single number. Number of surveys ~30 each quarter. Want to apply SPC limits to the quarterly, grouped number (ie % excellent among n = 30). Here's the problem, I think: 1. P-chart assumes rational subgroups that act independently of each other over time. 'Arbitrary' quarterly sub-groups are likely highly correlated since surveys are collected at random intervals throughout quarter and not at a specific point in time. 2. P-chart becomes ultra-sensitive as n gets large? I want to minimize the likelihood of a false positive (false alarm).





Is there such a thing as a moving average chart for proportions or attributes? Should I treat the individual proportions as variables data and use a MA or other chart? What are my options?

Best SPC control chart option?
You may want to generate the EWMA chart. It stands for Exponentially Weighted Moving Average control chart. Major statistical softwares would provide this option. For instance, SAS dose offer the procedure to construt EWMA chart, it is under SAS?QC component and the procedure it's called "PROC MACONTROL". You need to specified "EWMACHART" statement under PROC MACONTROL".


No comments:

Post a Comment