Profile cover photo
Profile photo
Gavin Simpson
261 followers
261 followers
About
Gavin's posts

Post has attachment
Modelling extremes using generalized additive models.

New blog post looking at how the new gevlss() family function in the #rstats pkg #mgcv can be used to fit statistical models to extremes data in the form of block maxima using the GEV. #R

Post has attachment
Simultaneous intervals for smooths revisited: correcting a silly mistake

New #rstats blog post in which I eat crow over an earlier post in which I claimed to generate a simultaneous confidence interval for a GAM but got it gob-smackingly wrong!

#R #statistics #gam #mgcv

Post has attachment
New on my blog: Pangaea and #R and open palaeo data. Featuring #rstats, the pangaear package, a GAM, #opendata, and secondary axes in #ggplot Something for everyone!
http://www.fromthebottomoftheheap.net/2016/12/16/pangaea-r-open-palaeo-data/

Post has attachment
In a new post I take a look at Achim Zeileis & Christian Kleiber's new paper & #rstats pkg, countreg, on rootograms for assessing the fit of count models. #R #statistics #counts #regression #models

For more, see Christian and Achim's paper, ArXived here: http://arxiv.org/abs/1605.01311

Post has attachment
LOESS Revisited, in which I take another look at why palaeolimnologists, inc me, abused LOESS & continue to do so http://www.fromthebottomoftheheap.net/2016/04/10/loess-revisited/ #science #palaeo #paleo #limnology #timeseries

Post has attachment
Soap-film smoothers & lake bathymetries. Here I look at the problem of finite area smoothing (smoothing over a region with a defined boundary) and at soap-film smoothers in particular. The motivating example is smoothing lake bathymetric data, from a consultancy project I was part of back in the UK. #rstats #r #spatial #smoothing #gam #additivemodels #statistics

Post has attachment
New post on my blog: Additive modelling global temperature time series: revisited. I revisit a post from 2011 and investigate evidence for a #globalwarming #hiatus using first derivatives of an additive model fitted to the HadCRUTv4 global mean temperature record. #rstats #r #climatechange #statistics #timeseries

Post has attachment
New post: Are some months warming more than others? Using an additive model with spline interactions, I discuss how to fit properly nested versions of these models in mgcv to facilitate comparing the additive and interaction models. And yes, some month are changing more than others, especially Oct and Nov, which warmed by >1°C over the last 100 years. http://bit.ly/1MMg26C #R #rstats #timeseries #statistics #climatechange

Post has attachment
New post: Climate change and spline interactions. Here I take a look at models that relax the additivity assumption and allow a smooth of one variable to interact smoothly with another. This is all illustrated via the Central England Temperature time series, and using our GAM we note substantial change in the within-year variation in temperatures, with some months warming by > 1°C http://bit.ly/1OlgI5j #R #rstats #timeseries #climatechange

Post has attachment
My aversion to pipes or why I'm not a fan, but would like to be, of the current pipe vogue in #rstats

Includes some excellent commentary too, especially the one from Gappy3000.

#r    #magrittr   #dply  
Wait while more posts are being loaded