ISSN:
1432-5225
Source:
Springer Online Journal Archives 1860-2000
Topics:
Agriculture, Forestry, Horticulture, Fishery, Domestic Science, Nutrition
,
Mechanical Engineering, Materials Science, Production Engineering, Mining and Metallurgy, Traffic Engineering, Precision Mechanics
Notes:
Summary Two variations of a class of permutation tests termed Multi-Response Permutation Procedures (MRPP1 and MRPP2) and the classical two-sample, two-sided t test were used to evaluate 72 data sets from tests on wood joints made with elastomeric construction adhesives. In all cases, the probability levels obtained from MRPP2 and the two-sample t test were nearly identical. This result stems from the fact that the test statistics of these two tests are theoretically equal. However, the underlying distributions of these two statistics are different. In several of the 72 comparisons, conflicting inferences about population differences were reached using MRPP1 and MRPP2. The results indicated that when the two data sets closely approximated a normal distribution and equal variances occurred, the MRPP2 (the permutation version of the t test) was the more optimal test. When validly-obtained extreme points were present, then the assumption of normality was not reasonable and MRPP1 was superior.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1007/BF00367732
Permalink