Erkki Liski - Academia.edu (original) (raw)
Papers by Erkki Liski
Evaluation of serum antibody response to a newly identified B-cell epitope in the minor nucleocapsid protein L2 of human papillomavirus type 16
Clinical and Diagnostic Virology, 1993
The aim of this work was to identify B-cell epitopes in the minor nucleocapsid (L2) protein of hu... more The aim of this work was to identify B-cell epitopes in the minor nucleocapsid (L2) protein of human papillomavirus (HPV) type 16 and characterization of allied antibody response. Serum samples of 513 individuals (323 women with various degrees of cervical atypia, 150 men and 40 small children) were available for the study. Synthetic peptides overlapping the L2 protein of HPV 16 twice were applied in ELISA for epitope scanning and antibody determination. An HPV 16 L2 derived dodecamer SGYIPANTTIPF (amino acids 391-402) proved to be the major B-cell epitope. Both IgA antipeptide antibody positivity (range 7-28%) and mean IgA antibody levels (range 13.2 EIU to 42.4 EIU, P < 0.05) increased with the degree of cervical atypia, whereas antipeptide IgG antibodies showed an opposite trend. During a 2-years follow-up significantly (P < 0.0005) decreasing IgA antibody levels to the SGYIPANTTIPF peptide were associated with regression of koilocytotic atypia. Analysis of anti-peptide IgA antibodies of 118 women with known HPV type revealed that a majority of positives had HPV 16/18 DNA. It was concluded that antibody response to the newly discovered peptide was partially type- and disease-specific. Our results also suggest an impairment of the IgG but not IgA class antibody response to HPV 16 in patients with persistent cervical HPV infection.
Topics in Optimal Design
Lecture Notes in Statistics, 2002
... Acknowledgments We would like to thank Professor AC Mukhopadhyay (Indian Statistical Institut... more ... Acknowledgments We would like to thank Professor AC Mukhopadhyay (Indian Statistical Institute, Calcutta), whose critical comments on the first few chapters of ... To my wife Sadhana and childrenMadhura and Pradip for allowing me to work without demanding much of my time ...
Prediction in Repeated-Measures Models with Engineering Applications
Technometrics, 1996
This article focuses on the problem of predicting future measurements on a statistical unit given... more This article focuses on the problem of predicting future measurements on a statistical unit given past measurements on the same and other similar units. We introduce a conditional predictor that uses the information contained in previous measurements. The prediction technique ...
Journal of Multivariate Analysis, 1996
In this paper, the Bayesian local influence approach is employed to diagnose the adequacy of the ... more In this paper, the Bayesian local influence approach is employed to diagnose the adequacy of the growth curve model with Rao's simple covariance structure, based on the Kullback Leibler divergence. The Bayesian Hessian matrices of the model are investigated in detail under an abstract perturbation scheme. For illustration, covariance-weighted perturbation is considered particularly and used to analyze two real-life biological data sets, which shows that the criteria presented in this article are useful in practice. 1996 Academic Press, Inc. Since Cook (1986) gave a general method for assessing the influence of local departure from assumption in certain statistical models, this approach, based on likelihood displacement, has played increasingly important roles in statistical diagnostics. This is due to the fact that all article no.
Information and Complexity in Statistical Modeling by Jorma Rissanen
International Statistical Review, 2007
International Statistical Review, 2009
Table of contents 1. The nature of research 10. Between-subjects factorial experiments: factors 2... more Table of contents 1. The nature of research 10. Between-subjects factorial experiments: factors 2. Principals of experimental design with more than two levels 3. The standard normal distribution: an amazing 11. Between-subjects factorial experiments: further approximation considerations 4. Tests for means from random samples 12. Within-subjects factors: one-way and 2 k factorial 5. Homogeneity and normality assumptions designs 6. The analysis of variance: one between-subjects 13. Within-subjects factors: general designs factor 14. Contrasts on binomial data: between-subject 7. Pairwise comparisons designs 8. Orthogonal, planned and unplanned comparisons 15. Debriefing 9. The 2 k between-subjects factorial experiment Appendix A. The method of least squares Appendix B. Statistical tables
Prediction of tree stems to improve efficiency in automatized harvesting of forests
Scandinavian Journal of Statistics, 1995
ABSTRACT. The problem of predicting future observations on a statistical unit given past measurem... more ABSTRACT. The problem of predicting future observations on a statistical unit given past measurements on the same and other similar units is frequently encountered in practical applications. When computer-based marking for bucking routines is used in a forest processor, it is usually ...
Acta Mathematicae Applicatae Sinica, 1996
A new condition for the antitonicity of the inverse of Hermitian matrices with respect to the L~w... more A new condition for the antitonicity of the inverse of Hermitian matrices with respect to the L~wner partial ordering is derived. The basic result is established by applying directly an elementary formula for a sum of matrices. K e y w o r d s . Hermitian, range-Hermitian, semidefinite matrices, Moore-Penrose inverse, sum of matrices
Acta Mathematicae Applicatae Sinica, 1999
In this paper we investigate properties of the power function of the generalized least squares F ... more In this paper we investigate properties of the power function of the generalized least squares F test for linear hypotheses under regression models with two-way error component model. The covariance structure of the model depends on the correlation coefficients m and ~ corresponding to the random effects. This model has been frequently applied to the analysis of panel data. In general, we show that the power is a monotonically increasing function of m (p2) in a region which is close to the m (,o2) axis, and a monotonically decreasing function of pl (,o2) in a region close to the ~ (m) axis.
Acta Mathematicae Applicatae Sinica, 1996
An inverse G of a given matrix A satisfying the condition GAG=G is known as a {2}-inverse of A. T... more An inverse G of a given matrix A satisfying the condition GAG=G is known as a {2}-inverse of A. This paper makes the {2}-inverse the starting point for studying the L~wner partial ordering of real nonnegative matrices. Simple basic properties of the {2}-inverses yield various extensions for the reverse ordering property, A> B if and only if B-x_>A -1 , of the positive definite matrices.
The paper considers a new model averaging method called weighted average least squares (WALS). Th... more The paper considers a new model averaging method called weighted average least squares (WALS). The method has good risk profile and its computational burden is light. The WALS technique can be easily applied to large data sets when the number of regressors is large. In the current paper the theory is used to compare the costs of hip fracture treatments between hospital districts in Finland.
The MDL model choice for linear regression
In this talk, we discuss the principle of Minimum Description Length (MDL) for problems of statis... more In this talk, we discuss the principle of Minimum Description Length (MDL) for problems of statistical modeling. By viewing models as a means of providing statistical descriptions of observed data, the comparison between competing models is based on the stochastic complexity (SC) of each description. The Normalized Maximum Likelihood (NML) form of the SC (Rissanen 1996) contains a component that may be interpreted as the parametric complexity of the model class. Once the SC for the data, relative to a class of suggested models, is calculated, it serves as a criterion for selecting the optimal model with the smallest SC. This is the MDL principle (Rissanen 1978, 1983) for model choice. If the parametric complexity of a model family is unbounded, then one must deviate from the clean definition of the SC. The most important example of this phenomenon is the Gaussian family. One approach to bound the parametric complexity is by constraining the sample space. We calculate the SC for the ...
In model selection one attempts to use the data to find a single ”winning” model, whereas with mo... more In model selection one attempts to use the data to find a single ”winning” model, whereas with model averaging (MA) one seeks a smooth compromise across a set of competing models. Most existing MA methods are based on estimation of single model weights using some appropriate criterion. The problem of selecting the best subset or subsets of predictor variables is a common challenge for a regression analyst. The number of candidate models may become huge and any approach based on estimation of all single weights may become computationally infeasible. Our approach is to convert estimation of model weights into estimation of shrinkage factors with trivial computational burden. We define the class of shrinkage estimators in view of MA and show that the estimators can be constructed using penalized least squares (LS) estimation by putting appropriate restrictions on the penalty function. The relationship between shrinkage and parameter penalization provides tools to build up computational...
Pitman Nearness, Distance Criterion and Optimal Regression Designs
Calcutta Statistical Association Bulletin
Pitman Nearness is a criterion for point estimation while the Distance Criterion is an optimality... more Pitman Nearness is a criterion for point estimation while the Distance Criterion is an optimality criterion. There is indeed a striking similarity between the two. In this paper we will present a variety of basic results towards estimation and ⁄ or prediction in the context of a regression model , based on the distance criterion .
A Percentile Regression Model for the Number of Errors in Group Conversation Tests
Optimal designs for prediction in random coefficient linear regression models
Influential Observations in the Generalize Analysis of Variance Model
Linear and Quadratic Growth Curve Models with lntraclass Covariance Structure and Related Optimal Designs
Metrika, 1995
On l�wner-ordering antitonicity of matrix inversion
Acta Math Appl Sin Engl Ser, 1996
Festschrift for Tarmo Pukkila on his 60th birthday. Papers presented at a specific session of the fifteenth international workshop on matrices and statistics, Uppsala, Sweden, June 13–June 17, 2006
Evaluation of serum antibody response to a newly identified B-cell epitope in the minor nucleocapsid protein L2 of human papillomavirus type 16
Clinical and Diagnostic Virology, 1993
The aim of this work was to identify B-cell epitopes in the minor nucleocapsid (L2) protein of hu... more The aim of this work was to identify B-cell epitopes in the minor nucleocapsid (L2) protein of human papillomavirus (HPV) type 16 and characterization of allied antibody response. Serum samples of 513 individuals (323 women with various degrees of cervical atypia, 150 men and 40 small children) were available for the study. Synthetic peptides overlapping the L2 protein of HPV 16 twice were applied in ELISA for epitope scanning and antibody determination. An HPV 16 L2 derived dodecamer SGYIPANTTIPF (amino acids 391-402) proved to be the major B-cell epitope. Both IgA antipeptide antibody positivity (range 7-28%) and mean IgA antibody levels (range 13.2 EIU to 42.4 EIU, P < 0.05) increased with the degree of cervical atypia, whereas antipeptide IgG antibodies showed an opposite trend. During a 2-years follow-up significantly (P < 0.0005) decreasing IgA antibody levels to the SGYIPANTTIPF peptide were associated with regression of koilocytotic atypia. Analysis of anti-peptide IgA antibodies of 118 women with known HPV type revealed that a majority of positives had HPV 16/18 DNA. It was concluded that antibody response to the newly discovered peptide was partially type- and disease-specific. Our results also suggest an impairment of the IgG but not IgA class antibody response to HPV 16 in patients with persistent cervical HPV infection.
Topics in Optimal Design
Lecture Notes in Statistics, 2002
... Acknowledgments We would like to thank Professor AC Mukhopadhyay (Indian Statistical Institut... more ... Acknowledgments We would like to thank Professor AC Mukhopadhyay (Indian Statistical Institute, Calcutta), whose critical comments on the first few chapters of ... To my wife Sadhana and childrenMadhura and Pradip for allowing me to work without demanding much of my time ...
Prediction in Repeated-Measures Models with Engineering Applications
Technometrics, 1996
This article focuses on the problem of predicting future measurements on a statistical unit given... more This article focuses on the problem of predicting future measurements on a statistical unit given past measurements on the same and other similar units. We introduce a conditional predictor that uses the information contained in previous measurements. The prediction technique ...
Journal of Multivariate Analysis, 1996
In this paper, the Bayesian local influence approach is employed to diagnose the adequacy of the ... more In this paper, the Bayesian local influence approach is employed to diagnose the adequacy of the growth curve model with Rao's simple covariance structure, based on the Kullback Leibler divergence. The Bayesian Hessian matrices of the model are investigated in detail under an abstract perturbation scheme. For illustration, covariance-weighted perturbation is considered particularly and used to analyze two real-life biological data sets, which shows that the criteria presented in this article are useful in practice. 1996 Academic Press, Inc. Since Cook (1986) gave a general method for assessing the influence of local departure from assumption in certain statistical models, this approach, based on likelihood displacement, has played increasingly important roles in statistical diagnostics. This is due to the fact that all article no.
Information and Complexity in Statistical Modeling by Jorma Rissanen
International Statistical Review, 2007
International Statistical Review, 2009
Table of contents 1. The nature of research 10. Between-subjects factorial experiments: factors 2... more Table of contents 1. The nature of research 10. Between-subjects factorial experiments: factors 2. Principals of experimental design with more than two levels 3. The standard normal distribution: an amazing 11. Between-subjects factorial experiments: further approximation considerations 4. Tests for means from random samples 12. Within-subjects factors: one-way and 2 k factorial 5. Homogeneity and normality assumptions designs 6. The analysis of variance: one between-subjects 13. Within-subjects factors: general designs factor 14. Contrasts on binomial data: between-subject 7. Pairwise comparisons designs 8. Orthogonal, planned and unplanned comparisons 15. Debriefing 9. The 2 k between-subjects factorial experiment Appendix A. The method of least squares Appendix B. Statistical tables
Prediction of tree stems to improve efficiency in automatized harvesting of forests
Scandinavian Journal of Statistics, 1995
ABSTRACT. The problem of predicting future observations on a statistical unit given past measurem... more ABSTRACT. The problem of predicting future observations on a statistical unit given past measurements on the same and other similar units is frequently encountered in practical applications. When computer-based marking for bucking routines is used in a forest processor, it is usually ...
Acta Mathematicae Applicatae Sinica, 1996
A new condition for the antitonicity of the inverse of Hermitian matrices with respect to the L~w... more A new condition for the antitonicity of the inverse of Hermitian matrices with respect to the L~wner partial ordering is derived. The basic result is established by applying directly an elementary formula for a sum of matrices. K e y w o r d s . Hermitian, range-Hermitian, semidefinite matrices, Moore-Penrose inverse, sum of matrices
Acta Mathematicae Applicatae Sinica, 1999
In this paper we investigate properties of the power function of the generalized least squares F ... more In this paper we investigate properties of the power function of the generalized least squares F test for linear hypotheses under regression models with two-way error component model. The covariance structure of the model depends on the correlation coefficients m and ~ corresponding to the random effects. This model has been frequently applied to the analysis of panel data. In general, we show that the power is a monotonically increasing function of m (p2) in a region which is close to the m (,o2) axis, and a monotonically decreasing function of pl (,o2) in a region close to the ~ (m) axis.
Acta Mathematicae Applicatae Sinica, 1996
An inverse G of a given matrix A satisfying the condition GAG=G is known as a {2}-inverse of A. T... more An inverse G of a given matrix A satisfying the condition GAG=G is known as a {2}-inverse of A. This paper makes the {2}-inverse the starting point for studying the L~wner partial ordering of real nonnegative matrices. Simple basic properties of the {2}-inverses yield various extensions for the reverse ordering property, A> B if and only if B-x_>A -1 , of the positive definite matrices.
The paper considers a new model averaging method called weighted average least squares (WALS). Th... more The paper considers a new model averaging method called weighted average least squares (WALS). The method has good risk profile and its computational burden is light. The WALS technique can be easily applied to large data sets when the number of regressors is large. In the current paper the theory is used to compare the costs of hip fracture treatments between hospital districts in Finland.
The MDL model choice for linear regression
In this talk, we discuss the principle of Minimum Description Length (MDL) for problems of statis... more In this talk, we discuss the principle of Minimum Description Length (MDL) for problems of statistical modeling. By viewing models as a means of providing statistical descriptions of observed data, the comparison between competing models is based on the stochastic complexity (SC) of each description. The Normalized Maximum Likelihood (NML) form of the SC (Rissanen 1996) contains a component that may be interpreted as the parametric complexity of the model class. Once the SC for the data, relative to a class of suggested models, is calculated, it serves as a criterion for selecting the optimal model with the smallest SC. This is the MDL principle (Rissanen 1978, 1983) for model choice. If the parametric complexity of a model family is unbounded, then one must deviate from the clean definition of the SC. The most important example of this phenomenon is the Gaussian family. One approach to bound the parametric complexity is by constraining the sample space. We calculate the SC for the ...
In model selection one attempts to use the data to find a single ”winning” model, whereas with mo... more In model selection one attempts to use the data to find a single ”winning” model, whereas with model averaging (MA) one seeks a smooth compromise across a set of competing models. Most existing MA methods are based on estimation of single model weights using some appropriate criterion. The problem of selecting the best subset or subsets of predictor variables is a common challenge for a regression analyst. The number of candidate models may become huge and any approach based on estimation of all single weights may become computationally infeasible. Our approach is to convert estimation of model weights into estimation of shrinkage factors with trivial computational burden. We define the class of shrinkage estimators in view of MA and show that the estimators can be constructed using penalized least squares (LS) estimation by putting appropriate restrictions on the penalty function. The relationship between shrinkage and parameter penalization provides tools to build up computational...
Pitman Nearness, Distance Criterion and Optimal Regression Designs
Calcutta Statistical Association Bulletin
Pitman Nearness is a criterion for point estimation while the Distance Criterion is an optimality... more Pitman Nearness is a criterion for point estimation while the Distance Criterion is an optimality criterion. There is indeed a striking similarity between the two. In this paper we will present a variety of basic results towards estimation and ⁄ or prediction in the context of a regression model , based on the distance criterion .
A Percentile Regression Model for the Number of Errors in Group Conversation Tests
Optimal designs for prediction in random coefficient linear regression models
Influential Observations in the Generalize Analysis of Variance Model
Linear and Quadratic Growth Curve Models with lntraclass Covariance Structure and Related Optimal Designs
Metrika, 1995
On l�wner-ordering antitonicity of matrix inversion
Acta Math Appl Sin Engl Ser, 1996
Festschrift for Tarmo Pukkila on his 60th birthday. Papers presented at a specific session of the fifteenth international workshop on matrices and statistics, Uppsala, Sweden, June 13–June 17, 2006