IN T R O D U C T IO N
I he conferences on M u ltiv a riate Statistical A nalysis have been organized every year since 1981.
T h eir goal has been focused on the latest achievem ents in the field o f theory o f m u ltiv ariate statistical analysis and its app licatio ns. T h e set o f papers we would like to present is the result o f the 20 In te rn a tio n a l C onference o f M u ltiv a riate Statistical A nalysis - M SA 2001. T h e conference was organised by the C h a ir o f Statistical M eth o d s o f the U niversity o f Ł ódź in co o p e ratio n w ith the Polish Statistical A ssociation. It w as held on N ovem ber 5-7, 2001, in the C onference C entre o f th e U niversity o f Ł ódź. T h e range o f subjects o f interest o f the C onference papers was fairly wide, from the theory of probability, through statistical m ethods to their applications in finance, in surance an d risk theory. T he p apers are divided in to three them atic groups:
1) S tatistical M eth o d s, 2) S tatistical Inference,
3) A pp licatio n o f Statistical M ethods.
T he first g ro u p includes articles with statistical m eth o d in a general sense, som e o f them are illustrated with exam ples.
Jerzy K orzeniew ski (M e th o d o f Colour Segmentation in Two Dimensional Images) considered som e selected algorithm s o f segm enting tw o dim ensional im ager. H e presents a new algorithm and com pariso n w ith o th er algorithm s.
W ojciech G a m ro t (A Two-phase Sampling Strategy f o r Estimating Multiple Mean Values in the Presence o f Nonresponse) fo rm u late th e o p tim izatio n problem lo r selecting sam ple and subsum able sizes m inim izing the expected value o f the co st of sim ultaneou s investigation o f m any a ttrib u te s o f p opulation .
I he second g ro u p o f papers deals w ith testing and estim ation procedures. C zeslaw D o m ań sk i (T he Estimate o f Power o f Random Tests B ased on Length o f Runs) fo rm u lated som e conclusion concerning pow er o f the m ost o ften ap p lied tests fo r ra n d o m n e ss based on th e len g th o f ru n s and restricted its to sta tio n a ry tw o state M ark o v chain.
K ry sty n a P ru sk a ( Testing the Identity o f Distributions o f two Discrete Random Variables) tested the identity o f distrib u tio n s fo r tw o un iv ariate and two bivariatc random variables. T h e pow er o f proposed tests is also analysed.
A gnieszka R ossa (Analysis o f Censored Life-tables with Covariates by means o f Log-linear M odels) p ro p o sed th e unbiased e stim a to r o f the survival p ro b a b ility and studied its asym ptotic properties.
Ja n u sz W yw iał ( Estimation o f Population Averages on the Basis o f a Vector o f Cluster Means) considered the estim ation o f a vector o f m ean values w hen the p o p u latio n is divided into clusters.
T o m asz Ż ąd ło (On M ean Square Error o f Synthetic Regression Estimator) presented the problem o f estim atio n of the to tal in a sm all d o m ain . H e proposed the synthetic regression estim ator.
M arcin Skibicki (On Maximization o f Estimation Accuracy in M ulti parameter Twostage Drawing) considered the problem o f d eterm in in g sam ple
sizes dow n in tw o stages so th a t the accuracy o f a n u m b er ol m eans estim ation w ould be m axim ized.
R obert Pietrzykowski and Wojciech Zieliński (A New Procedure o f Multiva riate Multiple Comparisons) presented the problem o f division the set of vector o f m eans into hom ogenous groups on the basis o f к sam ples. O n the basis of M o n te-C arlo sim ulations they show n th a t proposed m u ltip le co m p ariso n s procedure has high prob ab ility o f correct decision for the special cases.
Jan u sz W yw iał and T o m asz Ż ądło (On Som e Robust Against Outliers Predictor o f the Total Value in Sm all Domain) considered th e problem of prediction o f total value in dom ain based on simple regression superpopulation m odel. T hey show n the ro b u st estim ation against o u tlin crs o f regression fu n c tio n ’s p aram eter.
W iesław Pasiewicz an d W iesław W agner (Probabilistic M odel o f Ties in Multistage Decision Process) discussed the probability m o del on m u ltistag e decision process w ith p artic u la r em phasis on special case. A n idea o f im po rtance grap h ties is presented.
In the th ird g ro u p o f articles dealing w ith app licatio n s in finance, insurance, sp o rt, credit risk.
A d am Biela (Multidimensional Scaling in Economic Research) presented the possible directio n s o f applying m ultidim ensional scaling to social and econom ic analysis.
B ronisław C e ra n k a and M ałg o rzata G raczyk (Optimum Chemical Balance Weighing Designs Constructed f r o m the Incidence M atrices o f the Ternary Balanced Block Designs) studied the problem o f estim ation o f the weights of p objects in n w eighings using a chem ical balance design.
Jacek O lesinkiew icz and A n n a R u tk o w sk a-Z iark o (Application o j the W o l f s Algorithm to Constructing Effective Portfolios) presented the m eth o d o f co n stru c tin g the offective portfo lio w ith ap p licatio n o f W o lfs algorithm .
Sebastian Sitarz (Dynamic Programming with Returns in Random Variables Spaces) described a m odel o f dynam ic, discrete decision - m ak in g problem . He show n how this m odel could be applied for m ulti-stag e, m ulti-criteria decision m ak in g problem .
G rażyna 1 rzpiot (Preference Relations in Ranking Multivalued Alternatives in Finance Using Stochastic Dominance) propose a tw o-step p ro c ed u re for ranking alternatives u n d er am biguity. In decision situ atio n she co m pared m any alternatives.
A nna Szym ańska (Selected Statistical Methods o f Insurance Risk Assessment the Review o f Basic Methods) presented the m o st often applied statistical m ethod s for estim atin g insurance risk. T hese m eth o d s arc applied to the investigation o f the d istrib u tio n o f the nu m ber o f pay-off.
A gata S zczukocka (Selected Methods o f Credit Risk Evaluation) presented the selection m eth o d s o f credit risk: these m eth o d s allow to estim ate the risk o f single credit tra n sa c tio n and credit portfolio.
M ałg o rz ata M isztal ( The Use o j Som e Pattern Recognition Algorithms to Classify Patients Undergoing C A B G ) considered som e m eth o d s useful for predicting m o rtality risk after C o ro n ary A rtery B ypass G ra ftin g (C A B G ). She applied som e p a tte rn recognition algorithm s to classify p atien ts to one o f the risk subgroups.