ISSN:
1432-1343
Keywords:
Distance between groups
;
Location model
;
Mixed variables
;
Monte Carlo methods
;
Simulation
Source:
Springer Online Journal Archives 1860-2000
Topics:
Mathematics
Notes:
Abstract The location model is a useful tool in parametric analysis of mixed continuous and categorical variables. In this model, the continuous variables are assumed to follow different multivariate normal distributions for each possible combination of categorical variable values. Using this model, a distance between two populations involving mixed variables can be defined. To date, however, no distributional results have been available, against which to assess the outcomes of practical applications of this distance. The null distribution of estimated distance is therefore considered in this paper, for a range of possible situations. No explicit analytical expressions are derived for this distribution, but easily implementable Monte Carlo schemes are described. These are then applied to previously cited examples.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1007/BF01890125
Permalink