Software

 

Visualization

 

Predictive modeling

 

Dose-response modeling

 

Classification (https://www.molecularneuropathology.org/mnp)

 

Miscellaneous statistical topics

 

 

 

                               

Visualization

Clustering and visualization of mixed-type data

CluMix is an R package that provides utilities for clustering subjects and variables with mixed data types. The main feature is the creation of a mixed-data heatmap.

For details see:

Hummel M, Kopp-Schneider A. Clustering of samples and variables with mixed-type data. (Submitted) 

Visual analytics for the integrated analysis of microarray data

SEURAT provides interactive visualization capability for the integrated analysis of high-dimensional gene expression data. Gene expression data can be analyzed together with associated clinical data, array CGH (comparative genomic hybridization), SNP array (single nucleotide polymorphism) data and available gene annotations in an integrated manner.

For details see:

Gribov A*, Sill M*, Lück S, Rücker F, Döhner K, Bullinger L, Benner A, Unwin A (2010). SEURAT: visual analytics for the integrated analysis of microarray data. BMC Med Genomics;3:21. (* joint first authors). DOI: 10.1186/1755-8794-3-21

Biclustering via sparse singular value decomposition incorporating stability selection

s4vd is an addon package for the R-package biclust and provides implementations of the ssvd and s4vd algorithm to perform biclustering via sparse singular value decomposition with and without stability selection.

For details see:

Sill M, Kaiser S, Benner A and Kopp-Schneider A (2011). Robust biclustering by sparse singular value decomposition incorporating stability selection. Bioinformatics 27(15) 2089-2097. DOI:10.1093/bioinformatics/btr322

 

Predictive modeling

Extended inference for lasso and elastic-net regularized Cox and generalized linear models

The c060 package extends the popular R-package glmnet and provides additional functions particularly useful for high-dimensional risk prediction modelling, e.g. stability selection, estimation of prediction error (curves) and an efficient interval search algorithm for finding the optimal elastic net tuning parameter combination. Most functions offer improved computational efficiency through code parallelization.  

For details see:

Sill M, Hielscher T, Becker N, Zucknick M (2014). C060: Extended Inference with Lasso and Elastic-Net Regularized Cox and Generalized Linear Models. Journal of Statistical Software 62(5) 1-22. http://www.jstatsoft.org/v62/i05/

                                                     

Dose-response modeling

Design of dose-response studies

The WebApp DoseResponseDesigns calculates the optimal design for log-logistic, log-normal and Weibull functions. It also provides the D-efficiency of any given design compared with the optimal design.

For details see:

Holland-Letz T, Kopp-Schneider A. (2015). Optimal experimental designs for dose-response studies with continuous endpoints. Archives of Toxicology 89(11) 2059-2068.  DOI: 10.1007/s00204-014-1335-2

Analysis of dose-response studies

The WebApp MDRA performs dose-response analysis of multiple experiments. It allows for uploading of a csv-formatted data file for analysis. The four-parameter log-logistic model is used to fit dose-response data. Dose-response designs and data are visualized. Single experiments can be excluded from global analysis. Meta-analysis is used to average, e.g., EC50.

For details see:

Jiang, X, and Kopp-Schneider A. (2015). Statistical strategies for averaging EC50 from multiple dose-response experiments. Archives of Toxicology 89(11) 2119-2127. DOI 10.1007/s00204-014-1350-3

Jiang, X and Kopp-Schneider A. (2014). Summarizing EC50 estimates from multiple dose-response experiments: A comparison of a meta-analysis strategy to a mixed-effects model approach. Biometrical Journal 56(3): 493-512. DOI 10.1002/bimj.201300123

            

Classification

A molecular classifying algorithm for all types of tumors of the brain and its coverings as well as of tumors of peripheral nerves is available at MolecularNeuroPathology.org. Use of the classifier is currently password restricted to collaborating parties.                          

                              

Miscellaneous statistical topics

Cochran-Armitage Test for trend

The WebApp CATrend computes the one-sided p-values of the Cochran-Armitage trend test for the asymptotic and the exact conditional test. The Cochran-Armitage Test for trend is used in the analysis of 2 x k contingency tables  with k ordered categories. It compares the null hypothesis of equal proportions in all k categories to the alternative of ordered proportions. Details, also about numerical calculation can be found in the WebApp.

Sample size determination for diagnostic studies

The WebApp SampleSizeDiagnosticTest can be used to estimate the sample size for a study where the aim is to test whether the performance of a diagnostic test is sufficient in terms of false positive (specificity) and true positive fraction (sensitivity).

Analysis of semiparametric regression models

R package AdaptFitOS allows to fit additive regression models where covariates are permitted to have smooth effects of unknown functional form on a continuous response. Associated simultaneous confidence bands allow to assess the uncertainty of curve estimates and to study the statistical significance of certain features in a curve. The accompanied specification test provides statements about the significance of effects without relying on a parametric specification as in linear models.

For details see:

Wiesenfarth, M., Krivobokova, T., Klasen, S., Sperlich, S. (2012) Direct Simultaneous Inference in Additive Models and its Application to Model Undernutrition. Journal of the American Statistical Association, 107(500): 1286-1296. DOI: 10.1080/01621459.2012.682809

to top