Avviso di seminario

[marilena@pow2.sta.uniroma1.it (Marilena Barbieri)]

*******************************************************************

          DIPARTIMENTO DI STATISTICA, PROBABILITA' 
                 E STATISTICHE APPLICATE

            UNIVERSITA' "LA SAPIENZA" - ROMA

*******************************************************************

*******************************************************************

                        AVVISO DI SEMINARIO

*******************************************************************


   Lunedi' 16 dicembre, ore 11.30 - Aula VII 


      CLUSTER ANALYSIS WITH MIXED MODE DATA, MISSING VALUES
                   AND ERROR IN VARIABLES



                    Petros Dellaportas 
                    
	     Athens University of Economics
     	        Department of Statistics
	 	     Athens, Greece
	 	  



			Abstract


The classification of neolithic tools using cluster analysis 
enables archaeologists to understand the function of the tools 
and the technological and cultural conditions of the societies.
In this paper, Bayesian classification is adopted to analyse
data which raise the question of whether the observed
morphological-dimensional variability among the tools is related 
to their operational use.  The data present technical 
difficulties for the practitioners, such as the presence of 
mixed-mode data, missing data, and error in variables.	
These complications are overcome by employing a finite mixture 
model and Markov chain Monte Carlo methods.
The analysis uses prior information which expresses the 
archaeologist's belief that there are two tool groups similar
to contemporary adzes and axes.  The resulting mixing densities 
provide evidence that the morphological-dimensional variability
among tools is related to the existence of these two tool groups.

Keywords: Bayesian analysis; cluster analysis; error in variables;
Markov chain Monte Carlo; missing data; mixed mode data.