Category Archives: non-parametric statistics

Calculating Derivatives from Random Forests

Posted on 27 June 2020 by ecoquant

(Comment on prediction intervals for random forests, and links to a paper.) (Edits to repair smudges, 2020-06-28, about 0945 EDT. Closing comment, 2020-06-30, 1450 EDT.) There are lots of ways of learning about mathematical constructs, even about actual machines. One … Continue reading →

Posted in bridge to somewhere, Calculus, dependent data, dynamic generalized linear models, dynamical systems, ensemble methods, ensemble models, filtering, forecasting, hierarchical clustering, linear regression, model-free forecasting, Monte Carlo Statistical Methods, non-mechanistic modeling, non-parametric model, non-parametric statistics, numerical algorithms, prediction, R statistical programming language, random forests, regression, sampling, splines, statistical learning, statistical series, statistics, time derivatives, time series | Leave a comment

Simplistic and Dangerous Models

Posted on 27 April 2020 by ecoquant

Originally posted on Musings on Quantitative Palaeoecology:
A few weeks ago there were none. Three weeks ago, with an entirely inadequate search strategy, ten cases were found. Last Saturday there were 43! With three inaccurate data points, there is enough information…

Posted in Generalized Additive Models, non-parametric statistics, science, statistics | Leave a comment

Reanalysis of business visits from deployments of a mobile phone app

Posted on 20 February 2020 by ecoquant

Updated, 20th October 2020 This reports a reanalysis of data from the deployment of a mobile phone app, as reported in: M. Yauck, L.-P. Rivest, G. Rothman, “Capture-recapture methods for data on the activation of applications on mobile phones“, Journal … Continue reading →

Posted in Bayesian computational methods, biology, capture-mark-recapture, capture-recapture, Christian Robert, count data regression, cumulants, diffusion, diffusion processes, Ecological Society of America, ecology, epidemiology, experimental science, field research, Gibbs Sampling, Internet measurement, Jean-Michel Marin, linear regression, mark-recapture, mathematics, maximum likelihood, Monte Carlo Statistical Methods, multilist methods, multivariate statistics, non-mechanistic modeling, non-parametric statistics, numerics, open source scientific software, Pierre-Simon Laplace, population biology, population dynamics, quantitative biology, quantitative ecology, R, R statistical programming language, sampling, sampling algorithms, segmented package in R, statistical ecology, statistical models, statistical regression, statistical series, statistics, stepwise approximation, stochastic algorithms, surveys, V. M. R. Muggeo | 1 Comment

Procrustes tangent distance is better than SNCD

Posted on 24 February 2019 by ecoquant

I’ve written two posts here on using a Symmetrized Normalized Compression Divergence or SNCD for comparing time series. One introduced the SNCD and described its relationship to compression distance, and the other applied the SNCD to clustering days at a … Continue reading →

Posted in data science, dependent data, descriptive statistics, divergence measures, hydrology, Ian Dryden, information theoretic statistics, J.T.Kent, Kanti Mardia, non-parametric statistics, normalized compression divergence, quantitative ecology, R statistical programming language, spatial statistics, statistical series, time series | Leave a comment

Stream flow and P-splines: Using built-in estimates for smoothing

Posted on 28 January 2019 by ecoquant

Mother Brook in Dedham Massachusetts was the first man-made canal in the United States. Dug in 1639, it connects the Charles River at Dedham, to the Neponset River in the Hyde Park section of Boston. It was originally an important … Continue reading →

Posted in American Statistical Association, citizen data, citizen science, Clausius-Clapeyron equation, Commonwealth of Massachusetts, cross-validation, data science, dependent data, descriptive statistics, dynamic linear models, empirical likelihood, environment, flooding, floods, Grant Foster, hydrology, likelihood-free, meteorological models, model-free forecasting, non-mechanistic modeling, non-parametric, non-parametric model, non-parametric statistics, numerical algorithms, precipitation, quantitative ecology, statistical dependence, statistical series, stream flow, Tamino, the bootstrap, time series, water vapor | 2 Comments

Series, symmetrized Normalized Compressed Divergences and their logit transforms

Posted on 3 January 2019 by ecoquant

(Major update on 11th January 2019. Minor update on 16th January 2019.) On comparing things The idea of a calculating a distance between series for various purposes has received scholarly attention for quite some time. The most common application is … Continue reading →

Posted in Akaike Information Criterion, bridge to somewhere, computation, content-free inference, data science, descriptive statistics, divergence measures, engineering, George Sughihara, information theoretic statistics, likelihood-free, machine learning, mathematics, model comparison, model-free forecasting, multivariate statistics, non-mechanistic modeling, non-parametric statistics, numerical algorithms, statistics, theoretical physics, thermodynamics, time series | 4 Comments