Significant Help: Statistical Consulting and Case-Based Statistical Training
'Significant’ can be considered one of those not so intuitive terms in statistics. 'A significant effect' has the connotation of something meaningful while it only ensures you that some specified method of saying ‘yes’ or ‘no’ to your data is telling you ‘yes’. ‘Yes’ to an effect that might be the opposite of what you want it to be, ‘yes’ to an effect that might be so small that it is of no concern to your field and always a ‘yes’ that might be an improper ‘yes’.
Sometimes drawing conclusions from data is straightforward. But at other times, analytical methods are difficult and mistakes are easily made.
Significant Help's aim is to convey a sense of data by giving advice and customized training. Not only to yield the best results. But also, to give institutions and companies a notion of when statistics is easy, and when it is not. A distinction in all stages of the data process, designing experiments, the selection of modeling approaches and the actual analysis.
Significant is not always the same as important. In my work as a statistical consultant, I want to help businesses and institutions to decide what to extract from data that is important to them.
ALL-IN meta-analysis helps scientists to prioritize research, interpret findings and design new studies – so efficiently accumulate scientific knowledge. The ALL-IN approach is bottom-up, especially suited to combine studies that have no common (top-down) stopping rule.
"Scientists need to make sure that all relevant trials are in when they hedge their bets on future research."
This interview by Academic Positions gives a nice and short summary of my research at CWI.
My research is motivated by the 'replicability crisis' * in empirical science and the Reducing Research Waste movement in biomedical research ( http://rewardalliance.net/ ).
* This article sums up the 'crisis' quite well:
completed her Ph.D. research on ALL-IN meta-analysis at Centrum Wiskunde & Informatica ( CWI ), the national research institute for mathematics and computer science in the Netherlands, under the supervision of Prof. dr. Peter Grünwald and dr. Daniel Lakens . She now works as a consultant biostatistician at Amsterdam UMC, while still governing the Dutch Society for Statistics and Operations Research ( VVSOR ) as a treasurer.
For an overview of past projects, go to Projects.
T-test - Linear Regression - ANOVA - Logistic Regression - Generalized Linear Models - Robust Regression - Delta Method - Mixed Models - Longitudinal Modeling - Generalized Estimating Equations (GEE) - Bayesian Statistics - Discriminant Analysis - Optimal Scaling - Correspondence Analysis - Principal Component Analysis - Monte Carlo Statistics - Bootstrapping - Recommender System - Bayesian networks - K-means - Support Vector Machines - Multiple Testing Correction - High Dimensional Data Analysis - Ridge/Lasso - Splines - Cross-Validation - Random Forests - Bias/Variance Trade-off - Zero-Inflated Poisson Regression - Spatial/Temporal dependence - Finite Population Correction - Power Analysis - Randomized Incomplete Block Design - Survival Analysis - Censoring - Structural Equation Models - Item Response Theory - Imputation - EM Algorithm - Finite Mixture Models - Calibration Problem in Regression - R - SPSS - Weighted Contrasts - Stratified Design - Questionaire - Permutation Tests - Fixed/Common-Effect Meta-Analysis - Random Effect Meta-Analysis - Sequential Analysis - Bayes Factors - Bayesian T-test - Optional Stopping - Sequential Analysis - Meta-Analysis - Cummulative Meta-Analysis - Test Martingales - Uniformly Most Powerful Tests - Safe Tests