Comparing Discriminant Analysis and Logistic Regression Model as a Statistical Assessment Tools of Arsenic and Heavy Metal Contents in Cockles

Two statistical techniques; discriminant analysis (DA) and logistic regression model were used to analyze the concentration of arsenic and heavy metal contents in cockles (Anadara granosa) from two estuaries in the state of Penang, Malaysia. This study was undertaken to understand the interrelationship between different parameters and also to identify probable source component in order to explain the pollution status. Arsenic (As), chromium (cr), cadmium (cd), zinc (zn), copper (cu) and lead (pb) were analyzed using a graphite flame atomic absorption spectrophotometer (GF-AAS) whilst mercury (Hg) was analyzed using a cold vapor atomic absorption spectrophotometer (CV-AAS). Logistic regression model showed that only two explanatory variables Zn (p < 0.01) and Cd (p <0.05 ) exhibited significant effect to discriminate cockles in the two locations and responsible for large variation affording 77.5% correct assignation. On the other hand DA identified the same parameters Zn and Cd which are responsible in discriminating the two locations affording 72.5% correct assignation. Comparison between logistic regression model and DA exhibited that both techniques gave close results in discriminating the two locations.