Ideas: past (unfinished), current and future projects

I’ve got a headful of ideas that are driving me insane … (Bob Dylan, Maggie’s Farm)

These are all ideas that have floated through my head at one time or another that would be interesting to work on but that I simply haven’t gotten around to. Some of them could be appropriate for thesis (undergraduate, Master’s, Ph.D …) projects. Please contact me if any of them strikes you as interesting.

Epidemics

• use multivariate autoregressive state-space models (MARSS, e.g. from Eli Holmes’s MARSS package to explore the dynamics either of multi-species pathogens (e.g. West Nile virus, Batrachochytrium dendrobatidis) or of multiple pathogens in a single host (e.g. data from David Earn on London Bills of Mortality)
• explore the degree to which initial epidemic curves seem to follow power-law, rather than exponential, trajectories? (inspired by Australian AIDS data from Dobson and Barnett)
• use monotonic splines and/or appropriately constrained Gaussian processes to fit “semi-mechanistic” epidemic models, i.e. models where transmission is a smoothly increasing function of \(S\) and \(I\) but need not follow any particular functional form (Forkutza)
• modeling targeted testing and effects of time-variable testing on test positivity (cf Campbell et al. (2021) and these notes)

Functional responses

• Review/polishing of approaches for FR modeling incorporating depletion, multi-species, etc. etc.. Old notes here
• Push forward ideas from Okuyama and Bolker?
• semi-mechanistic FRs (see last bullet point under “Epidemics”)
  • related to questions about model sensitivity (see notes from 2019 BIRS meeting; papers by Wolkowicz et al. on structural sensitivity of FR models (Eastman 2017; Seo and Wolkowicz 2018); also Aldebert and Stouffer (2018)

Variance dynamics/autocorrelated growth

• Extend growth variance model to handle other components (size-dependent growth and mortality)
• More general growth analysis: Peacor data?

Related: Brooks, McCoy, and Bolker (2013)

Evolution of virulence

• better analytical handle on “simple” JRSI approach; can we get anywhere with characterizing (approximately) the peak height/timing?
• general-purpose Mendelian genetics simulator?
• finish paper(s) with Morgan Kain on general virulence dynamics models

Related: Bolker, Nanda, and Shah (2010)

Common principal components

• test, polish, document CPC code written by Alan Wong
• work out the multi-group BPANOVA (back-projection ANOVA) method
• review of quadratic vs geometric approaches to size-correction

Related: McCoy et al. (2006)

Spatial/phylogenetic (G)LM(M)s

• compare existing methods: geoRglm, INLA, gee, AD Model Builder, nlme, … (Keitt 2002; Dormann et al. 2007)
• Revamp Ives and Godfray (2006) with a proper coevolutionary model

Related: flexLambda branch of lme4

General-purpose MLE software

• push qzmle forward

General population dynamics

• coral dynamics (AIMS)

Mixed models

• further investigation of Zhang et al. (2011): comparative performance of different GLMM approaches; e.g. bambi/PyMC (Python), rstanarm, brms, greta (PyTorch), MixedModels.jl (Julia), glmmTMB, lme4, … also SAS PROC MIXED??
• improve Julia-to-R connections
• Thousands of potential improvements to lme4 and glmmTMB

General spatial dynamics

• Develop a general-purpose point-process spatial simulator (for use with emonk that implements a wide set of spatial transition rules. Transition rules should be either parsed into a C function that can be compiled with the rest of the structural code, or parsed into a fairly rapidly executable set of instructions (à la Bard Ermentrout’s xppaut and xtc). Facilities for incorporating or generating maps of environmental heterogeneity? More ambitious: parallel patch and IPS models, for comparison (alternately, extend s3 to allow more quantitative analysis and make it more of a research tool)

Dispersal

• explore power, bias, etc. of inverse fitting approaches to estimation of (biological/seed) dispersal kernels, especially with multiple sites

Misc.

(as though this whole list couldn’t be lumped under “miscellaneous” …)

• objective Bayesian “tuning” of forest models/other large-scale models with data from multiple scales
• scale-dependent regression (see here)
• implement small-sample methods from Das, Dhar, and Pradhan (2018) ?
• Shiny apps for BIO 4AE3 “ecology and evolution of infectious disease”; pop cycles, Red Queen, evolution of resistance/tolerance, etc etc etc

More unfinished projects/ideas

This is an embarrassingly long list of projects I’ve worked on over the years that could be revisited. They’re either unpublished or unfinished/want to be revisited or both.

• reef fish (see old talk)
  • Previously attempted brute-force Gibbs sampling, but complexity became overwhelming; I would consider revisiting this using Approximate Bayesian Computation and/or sequential Monte Carlo (e.g. using the pomp package; can we build in random effects/smoothing?)
• turtle return times
• mosquito population estimation (J. Walton)
• improve/polish mixed-stock analysis machinery
• corHMM + TMB (in progress, L. Eckert)
• revisit spatial deconvolution (Moran effect, thunderstorms & lemmings, pines (ESA talk, progress report, Seabloom et al. (2005), …) (A. Berk; in progress, J Freeman)
• revisit journal cross-citation study (A. Berk)
• multi-factorial plague: revisit with Stan (in progress J. Pekos)
• extend lunchbox (M. Li): more gamma-white noise/moment-matching etc. to allow use in TMB/Stan, with latent time variation
• longitudinal community dynamics via latent/reduced-rank methods, especially microbiomes (in progress, M. Agronah?)
• finish cricket/Duckworth-Lewis project (Bhatti)

Software

Lots …

• clean up mixstock package
• open-source bibliometrics using OpenAlex or Semantic Scholar
  • https://hachyderm.io/@ethanwhite/110137220689858030
  • https://ecoevo.social/@noamross/110136395357840168
• penalized wavelets à la Wand and Ormerod (2011)