Running with Scissors

Presenter: Paul Sheehy, Principal, Paul Sheehy Consulting, Brewer, ME, USA

Keywords: Statistics, Assumptions, Common Errors

Industry: Education/Training, Manufacturing, Service

Level: Intermediate


ABSTRACT

Give people a powerful tool (statistics) with insufficient training, and you likely have a case of somebody “running with scissors.” This talk will cover some of the more common statistical mistakes including:

  • Misapplication of the normality assumption and normality testing with a discussion of which statistical tools or methods are sensitive to the assumption, which are moderately affected and which are robust to it. Additionally, a method to validate normality with large samples will be demonstrated.
  • Questions regarding the equal variance assumption for the 2-Sample t test and One-Way ANOVA. This will include the use of the Welch’s test.
  • Criticality of rational subgroups in control charting and capability.
  • In measurement systems, the issues arising from the use of the small number of parts in a typical Gage R&R study to estimate the variation of the process which, if unverified, may result in a seriously incorrect denominator when calculating the % gage R&R. There will be a discussion on the value of using a historical standard deviation as a check.
  • Issues that can seriously impact the ability of a Capability Analysis to predict the future performance of the process. Discussion will include the effects due to insufficient data, insufficient time, subgroup problems, and lack of stability. The concept of a Capability Snapshot will be presented.