ABSTRACT

Hypothesis rejection in an analysis of variance implies that two or more group means under test are significantly different. The significant differences can be localized using a priori tests (i.e., orthogonal comparisons) only if prior knowledge has allowed the experiment to be designed in a manner amenable to such analysis. In some geologic work this is not possible; in such cases multiple comparisons testing is then performed to determine exactly where the significant differences lie. Failure to carry out this a posteriori testing could limit the interpretation of the original analysis of variance to vague conclusions concerning the means.

Multiple comparisons methods are also useful for grouping objects, thus allowing an investigator to find, in addition to all contrasts, underlying groupings to which geologic meaning can sometimes be attached. Because multiple comparisons methods are not frequently employed by geologists, the present study illustrates three of these methods as applied to geological data.

An application of sum of squares simultaneous test procedure (SS-STP), range simultaneous test procedure (Range-STP), and Student-Newman-Keuls (SNK) methods to 5 sampling transects across sand waves in Chesapeake Bay entrance has shed light on observed textural variations. Using SS-STP method, crest sediment has been found in many cases to be significantly different from trough sediment in mean grain size. Groups that result from an application of Range-STP and SNK methods show correspondence between mean grain size and sand wave topography; crests and troughs can each be frequently classified as statistically similar groups.