50 



3.10.4 Discriminant Statistical Analysis 



The final statistical analysis was performed again with the data used to group the 

 postcap samples into three pre-determined layers. Using SPSS® Professional Statistics 6.1, 

 we performed a discriminant statistical analysis on the mineralogy and microfossil results 

 from the core samples. Discriminant statistics is a multivariate program that identifies and 

 forms linear combinations of independent variables which is then used to classify the 

 samples into groups (Norusis 1994). Because the groups were pre-determined, the success 

 of classification provides information on the differences between or similarities within the 

 groups. 



The mineralogy data were quantified the same way as previously described for the 

 clustering analysis. In the SPSS package, the microfossil data were grouped into five 

 categories that were used to display and interpret the data: freshwater thecamoebians, 

 marsh foraminifera, mudflat foraminifera, shelf agglutinated foraminifera, and shelf 

 calcareous foraminifera. The relative abundance of the five groups of species were 

 calculated for individual samples. Each sample was then grouped as ambient, pseudo- 

 UDM, or CDM based on the depth of the sample with respect to the visual core 

 descriptions and on microfossil content. The layer classification was the same as for the 

 PRIMER statistical analyses. 



The program calculated group means, standard deviation, and discriminant scores. 

 To measure the degree of association between the scores and the groups, the discriminant 

 scores were graphed according to two canonical discriminant functions. The canonical 

 functions represent the ordination axes that best separate the known groups. The program 

 determined the percentage of samples successfully classified into the pre-determined groups. 



The Portland Disposal Site Capping Demonstration Project, 1995-1997 



