In recent years, a surprising number of scientific results have failed to hold up to continued scrutiny. Part of this ‘replicability crisis’ may be caused by practices that ignore the assumptions of traditional (frequentist) statistical methods (John, Loewenstein, and Prelec 2012). One of these assumptions is that the experimental protocol should ...

Introduction In the last decade, there have been spectacular advances on the practical side of machine learning. One of the most impressive may be the success of Generative Adversarial Networks (GANs) for image generation (Goodfellow et al. 2014)....

Introduction Let us start by considering the optimal rejection policy for a single hypothesis testing problem. There are three elements to the problem. The objective: to maximize the power to reject the null hypothesis; The constraint: to control ...

Abstract We consider the problem of learning a function defined on a compact domain, using linear combinations of a large number of “bump-like” components (neurons). This idea lies at the core of a variety of methods from two-layer neural networks...

In the toolbox of most scientists over the past century, there have been few methods as powerful and as versatile as linear regression. The introduction of the generalized linear model (GLM) framework in the 1970’s extended the inferential and pred...

Summary We propose a higher order targeted maximum likelihood estimation (TMLE) that only relies on a sequentially and recursively defined set of data-adaptive fluctuations. Without the need to assume the often too stringent higher order pathwise d...

Compositional data are characterized by the fact that the relevant information is contained not necessarily in the absolute values but rather in the relative proportions between particular components. As an example, take household expenditures for ...

Correlations and negative bias We assume that you are quite familiar with the following problem. Consider a data set where the information is expressed in percentages or proportions. An example are household expenditures, given as average amounts (...

Abstract Binary classification is one of the cornerstones of modern data science, but, until recently, our understanding of classical methods such as the k-nn algorithm was limited to settings where feature vectors were compactly supported. Based o...

We look at the problem of learning latent structure between two blocks of data through the partial least squares (PLS) approach. These methods include approaches for supervised and unsupervised statistical learning. We review these methods and prese...

I’ve kindly been invited to share a few words about a recent paper my colleagues and I published in Bayesian Analysis: “Bayesian Regression Tree Models for Causal Inference: Regularization, Confounding, and Heterogeneous Effects”. In that paper, we motivate and describe a method that we call Bayesian causal forests (BCF), ...

Abstract Co-clustering techniques, which group observations and features simultaneously, have proven to be efficient in summarising data sets. They exploit the dualism between rows and columns and the data set is summarized in blocks (the crossing of a row-cluster and a column-cluster). However, in the case of mixed data sets (...

We address the problem of causal structure learning in the presence of hidden variables. Given a target variable and a vector of covariates, we are trying to infer the set of observable causal parents of the target variable. There are many good reasons...

Introduction New statistical process control (SPC) methods have to be developed in order to handle more and more complex data, which are available because of the advent of new data acquisition technologies. In particular, in many practical situations t...

The Hungarian Central Statistical Office uses online cash register data for the measurement of retail trade and food services. The data source allows the statistical office to considerably cut administrative burden of data providers without quality los...

We first started making risk calculators with the website, http://nomograms.org. There are several here, and they are patient friendly. Later, we launched rcalc.ccf.org, which has many more risk calculators, although the intended audience is the ...