Nick Clark 💹
banner
njclark.bsky.social
Nick Clark 💹
@njclark.bsky.social
Amplify Research Fellow at University of Queensland. Using #forecasting in #rstats to understand ecological responses to change. Author of the {mvgam} R 📦 for Bayesian multivariate modeling.
Blog 💬: https://ecogambler.netlify.app/
Views are my own
Compositional data needs models that respect its constraints. I show how Dirichlet regression with GPs captures nonlinear environmental effects on plant communities while respecting the dependence of the data #rstats #dataviz ecogambler.netlify.app/blog/plant-c...
October 21, 2025 at 8:19 AM
Want to move beyond simple forecasting algorithms for modelling real world time series? The {mvgam} #rstats 📦 has you covered nicholasjclark.github.io/mvgam/
May 14, 2025 at 7:11 PM
Great thanks for the helpful tips! I think this is getting there now :)
April 4, 2025 at 5:47 AM
Some new additions in the {mvgam} #rstats 📦 now make it easier to visualise the complex associations among species that can be inferred from latent variable models nicholasjclark.github.io/mvgam/refere...
April 3, 2025 at 9:03 AM
Thanks for your suggestions. I incorporated most of them, but still needs refinement
April 3, 2025 at 8:51 AM
March 12, 2025 at 2:35 AM
I'm pleased with how these Impulse Response Function plots have turned out. The {mvgam} #rstats 📦 makes it easy to fit multivariate time series models and compute these, even with non-Gaussian data github.com/nicholasjcla...
March 6, 2025 at 6:28 AM
The latest release of the {mvgam} #rstats 📦 has hit CRAN. Plenty of exciting new features including Joint Species Distribution Models, support for the full range of Gaussian Process kernels available in the {brms} 📦 and plenty more nicholasjclark.github.io/mvgam/news/i...
February 19, 2025 at 10:09 PM
We found that hierarchical dynamic models that allow species to interact over time performed better than simpler models, both during in-sample hindcast and out-of-sample forecast scenarios
February 19, 2025 at 10:04 PM
For those in Southeast Queensland/Northern NSW, don't miss out on some upcoming #rstats workshops on regression modeling, spatiotemporal models and multivariate statistics mathmarecol.github.io/RWorkshop/. Running in-person in early February
January 21, 2025 at 9:35 PM
Love it. I implemeted something similar for a recent blog post (ecogambler.netlify.app/blog/vector-...), using some suggestions from @mjskay.com
January 13, 2025 at 4:23 AM
Looking forward to delivering two workshops on time series and forecasting with {mvgam} for #ESAus2024 #rstats. It'll be a busy summer of workshop for me 😄 ecogambler.netlify.app/talk/
December 8, 2024 at 6:16 AM
Proud of this work-in-progress function in my {mvgam}📦(nicholasjclark.github.io/mvgam/refere...) to auto-generate a fully referenced methods scaffold that people can use as a template for describing their models. Would love to hear of similar functionalities from other packages #rstats
December 4, 2024 at 5:17 AM
Just learned about the "Enhances" field in #rstats📦 development and it is👍. Have added support for draw() methods from {mvgam} models if users have
@gsimpson.bsky.social's {gratia} installed. This automatically supports {mvgam} specials such as monotonic, GPs and dynamic functions
July 25, 2024 at 1:05 AM
Saw this on X and had to crosspost: Peer review encapsulated
May 12, 2024 at 9:24 PM
Very pleased to announce that the {mvgam} #rstats package, which fits State-Space Dynamic GAMs to time series, has been released to CRAN. Here is a short blog post to describe current features and future plans: ecogambler.netlify.app/blog/mvgam-o...
May 7, 2024 at 11:51 PM
Looking forward to a second rendition of this course on forecasting with GAMs in #rstats. If you are interested, have a look at some of my webinars to get a feel for what the course covers ecogambler.netlify.app/talk/
April 13, 2024 at 5:57 AM
A recent post on Cross Validated motivated me to write a quick blog post on leveraging the power of the {marginaleffects} 📦 to interpret effects from GAMs fitted in the {mgcv} 📦. Thanks @vincentab.bsky.social for helpful comments #rstats ecogambler.netlify.app/blog/interpr...
March 29, 2024 at 3:42 AM
A recent post on StackOverflow about difficulties setting up by-terms in #mgcv distributed lag models (stackoverflow.com/questions/75...) reminded me of a case study I wrote up last year. Have since added it to my very small blog here: ecogambler.netlify.app/blog/distrib... #rstats
March 14, 2024 at 6:03 AM
The mvgam📦in #rstats has evolved hugely since the last rendition of this course. Too much to cover in one workshop but see the website for details nicholasjclark.github.io/mvgam/
March 6, 2024 at 11:51 PM
Reminder that I'll be running a second edition of my Ecological forecasting in R workshop in May this year. Learn to use #mgcv, #mvgam and #brms to analyse complex ecological time series and produce meaningful insights #rstats 🌍 www.physalia-courses.org/courses-work...
March 6, 2024 at 11:48 PM
We find support for complex, delayed covariate effects and smooth temporal dynamics. Forecast uncertainties are controlled by both components, suggesting there is room for improvement (for example, using multispecies models such as here: ecoevorxiv.org/repository/v...
February 25, 2024 at 11:51 PM
We also follow
@betanalpha.bsky.social's guidance to develop bespoke checking functions to replace generic criteria (AIC, DIC) for targeted model expansion
February 25, 2024 at 11:50 PM
Exciting new work in #ecologicalforecasting from my PhD student (authors.elsevier.com/sd/article/S.... We use Bayesian Dynamic GAMs to model time series of captures for desert rodents. A short thread:
February 25, 2024 at 11:48 PM
February 14, 2024 at 7:59 AM