Deleted Residual
Process of calculating residuals in which the influence of each observation is removed when calculating its residual. This is accomplished by omitting the ith observation from regression equation used to calculating it predicted value.
 
 
     
 
  Deviance
A measure of fit of a statistical model. The deviance is twice (log-likelihood of the best model minus log-likelihood of the current model). The best model can be the true model or an exact fit (often called a saturated model).
 
 
     
 
  DFBETA
Measure of the change in a regression coefficient when an observation is omitted from the regression analysis. The value DFBETA is in terms of the coefficient itself; a standardized form(SDEBETA) is also available. No threshold limit can be established for DFBETA. Although the researcher can look for values substantially different from the remaining observations to assess potential influence. The SDDFBETA values are scaled by their standard errors. Thus supporting the rational for cutoffs of 1 or 2, corresponding to confidence levels of 0.10 or 0.05, respectively.
 
 
     
 
  Diagnostic Paradigm
In the terminology of Dawid (1976), modelling the conditional distribution of the class C given the features X.
 
 
     
 
  Dirichlet Distribution
A distribution over probability distributions () on K classes. Its density (Berger, 1985, p.561) is, for and which has mean at (), and is increasingly concentrated as () increases.
 
 
     
 
  Dirichlet Tessellation
Given a set of points in associate with each those points of to which it is nearest. This defines a tile, and the tiles partition the space, Also known as Voronoi or Thiessen polygons in Preparata & Shamos (1985) give algorithmic details.
 
 
     
 
  Dissimilarity
A measure of the dissimilarity of two examples based on their features. Must be non-negative and symmetric.