Viktor Witkovsky
Mailing Address:
RNDr. Viktor Witkovsky, CSc.
Department of Theoretical Methods
Institute of Measurement Science
Slovak Academy of Sciences
Dubravska cesta 9
84219 Bratislava
Slovak Republic
E-mail: Witkovsky@savba.sk
Current research interests:
My current research interest is in exact and approximate statistical methods for
small sample inference. In particular, I am interested in computational aspects of
making statistical inference
on the parameters in fixed and random effects models. Of particular interest is
the one-way classification model with heteroscedastic errors which has direct
application in the interlaboratory studies, clinical trials and meta-analysis.
Other closely related research interests:
Binomial Proportions. An algorithm for computing the exact and the mid-P confidence intervals for difference of binomial proportions was suggested. The
method is based on combining the so called consonance distributions for the binomial proportions. The calculation of the lower and the upper bound of the interval estimate does not depend on the nuisance parameter and requires numerical integration of well behaved functions. See the Technical Report.
The Behrens-Fisher Problem. The methods to calculate the
p-values required for deriving the conservative joint confidence interval estimates for
the pairwise mean differences, refered to as the generalized Scheffe intervals were suggested. Further, the corresponding tables with critical values for simultaneous comparisons of the mean differences of k normal populations with unequal variances based on independent random samples, including very small sample sizes, were calculated. See the Technical Report.
MATLAB Algorithm Mixed.m. The MATLAB algorithm was developed
to estimate
the parameters of the mixed linear model by using the Henderson's MMEs (mixed model equations) algorithm.
The algorithm computes the BLUE (or the two-stage GLSE), the BLUP
(or the EBLUP), and the ML, REML, MINQE(I), or MINQE(U,I) of the
fixed effects, the random effects, and the variance components, respectively,
in the mixed linear model, together with the generalized inverse of
the MMEs coeffcient matrix and the Fisher information matrix for the
estimated variance components. Get the m-file from MATLAB Central > File Exchange > Statistics and Probability > mixed and/or the Technical Report.
The research has been supported by the following Grants from Slovak Grant Agency VEGA:
Statistical models and methods. Grant VEGA 1/4196/97, Jan 1997 - Dec 1999, Principal Investigator: Prof. RNDr. Andrej Pázman, DrSc.
See the final report (in Slovak).
Statistical models and methods II. Grant VEGA 1/7295/20, Jan 2000 - Dec 2002, Principal Investigator: Prof. RNDr. Andrej Pázman, DrSc.
See the final report (in Slovak).
VW-Publications ¤ VW-Algorithms-Software ¤
Journal on Measurement Science:
Measurement Science Review ¤
Current Conferences:
ProbaStat 2006 ¤
Organized Conferences (History):
ProbaStat 2002 ¤
MEASUREMENT 2003 ¤
MEASUREMENT 2005 ¤
Other Links:
Statistics from Wikipedia
Powered by counter.bloke.com