Loading…
2015 Spring Symposium on Undergraduate Research and Community... has ended
Wednesday, April 22 • 8:30am - 10:30am
A Methodology for Appropriate Testing When Data is Heterogeneous Using EXCEL

Sign up or log in to save this to your schedule and see who's attending!

A Methodology for Appropriate Testing When Data is Heterogeneous was originally published and copy written in the mid-1990s in Turbo Pascal using the a16 bit operating system. While working on an ergonomic dissertation (Yearout, 1987), the perceptual data was determined to be heterogeneous and not normal. Dr. Milliken and Johnson, the authors of Analysis of Messy Data Volume I: Designed Experiments (1989),advised that Satterthwaite’s Approximation with Bonferroni’s Adjustment to correct for pairwise error be used to analyze the heterogeneous data. This technique of applying linear combinations with adjusted degrees of freedom allowed the use of t-Table criteria to make group comparisons without using standard nonparametric techniques. Thus data with unequal variances and unequal sample sizes could be analyzed without losing valuable information. Variances to the 4th power were so large that they could not be reentered into basic calculators. The solution was to develop an original software package which was written in Turbo Pascal on a 7 ¼ inch disk 16 bit operating system. Current operating systems of 32 and 64 bits and more efficient programming languages have made the software obsolete and unusable. Using the old system could result in many returns be either incorrect or the system would terminate. The purpose of this research was to develop a spreadsheet algorithm with multiple interactive EXCEL worksheets that will efficiently apply Satterthwaite’s Approximation with Bonferroni’s Adjustment to solve the messy data problem. To insure that the pedagogy is accurate the resulting package was successfully test tested in the classroom with academically diverse students. A comparison between this technique and EXCEL’s Add-Ins Analysis Toolbar for a t-test Two-Sample Assuming Unequal Variances was conducted using several different data sets. The results of this comparison was that the EXCEL Add-Ins returned incorrect significant differences. Engineers, ergonomists, and social scientists at home and abroad would find the developed program very useful. A major benefit is that spreadsheet algorithms will continue to be current regardless of evolving operating systems’ status


Wednesday April 22, 2015 8:30am - 10:30am
Wilma Sherrill Center Concourse