In 1957, linguist J. R. Firth observed that 'you shall know a word by the company it keeps'. That principle — words that co-occur share meaning — is the foundation on which all of generative AI was built, from early Latent Semantic Analysis to today's trillion-parameter Transformers. This post traces the lineage with three interactive LSA-to-PCA visualisations in R (Reuters newswire, State of the Union addresses and IMDB reviews), showing where simple co-occurrence models succeed, where they fail and why scale alone turned a modest insight into the technology behind ChatGPT. It then examines why LLMs are optimised for fluency rather than truth — hallucinations are a structural consequence, not a bug to be patched — and argues that careful prompt engineering is the best tool we have for steering a fundamentally heuristic machine.
Electroencephalography (EEG) has become a cornerstone for understanding the intricate workings of the human brain in the field of neuroscience. However, EEG software and hardware come with their own set of constraints, particularly in the management of markers, also known as triggers. This article aims to shed light on these limitations and future prospects of marker management in EEG studies, while also introducing R functions that can help deal with vmrk files from BrainVision.
Extension of the rscopus R package with functions that manage search quotas, retrieve DOIs for reference managers, search for additional DOIs, compare publication counts across topics, and visualize bibliometric comparisons over time.
Frequently asked questions about mixed-effects models, covering the necessity of random slopes, appropriate p-value calculation methods, parallelization limitations, convergence issues, and optimizer selection.
In the fast-paced world of scientific research, establishing minimum standards for the creation of research materials is essential. Whether it's stimuli, custom software for data collection, or scripts for statistical analysis, the quality and transparency of these materials significantly impact the reproducibility and credibility of research. This blog post explores the importance of adhering to FAIR (Findable, Accessible, Interoperable, Reusable) principles, and offers practical examples for researchers, with a focus on the cognitive sciences.
An R script for preprocessing frequency list data from the Norwegian Web as Corpus (NoWaC), including instructions for downloading and preparing the corpus data.
Custom functions that extend sjPlot to discretise continuous variables in interaction plots into equally-sized intervals (deciles or sextiles) that include minimum and maximum values, with legend counts showing sample sizes per level.
Whereas the direction of main effects can be interpreted from the sign of the estimate, the interpretation of interaction effects often requires plots. This task is facilitated by the R package sjPlot. For instance, using the plot_model function, I plotted the interaction between a continuous variable and a categorical variable. The categorical variable was passed to the fill argument of plot_model.
library(lme4) #> Loading required package: Matrix library(sjPlot) #> Install package "strengejacke" from GitHub (`devtools::install_github("strengejacke/strengejacke")`) to load all sj-packages at once!
A custom R function to create ggplot2 visualizations of fixed effects from models refitted with multiple optimizers using lme4's allFit function, enabling visual assessment of convergence validity in mixed-effects models.
Implementation of fuzzy string matching in R using fuzzyjoin::stringdist_join with controlled fuzziness via max_dist argument, demonstrated with food name matching and frequency-based code selection.
When a model has struggled to find enough information in the data to account for every predictor---especially for every random effect---, convergence warnings appear (Brauer & Curtin, 2018; Singmann & Kellen, 2019). In this article, I review the issue of convergence before presenting a new plotting function in R that facilitates the visualisation of the fixed effects fitted by different optimization algorithms (also dubbed optimizers).
Here I share the format applied to tables presenting the results of Bayesian models in Bernabeu (2022). The sample table presents a mixed-effects model that was fitted using the R package 'brms' (Bürkner et al., 2022).
Here I share the format applied to tables presenting the results of frequentist models in Bernabeu (2022). The sample table presents a mixed-effects model that was fitted using the R package 'lmerTest' (Kuznetsova et al., 2022).
Whereas the direction of main effects can be interpreted from the sign of the estimate, the interpretation of interaction effects often requires plots. This task is facilitated by the R package sjPlot (Lüdecke, 2022). In Bernabeu (2022), the sjPlot function called plot_model served as the basis for the creation of some custom functions. One of these functions is alias_interaction_plot, which allows the plotting of interactions between a continuous variable and a categorical variable.
Custom functions extending sjPlot for plotting interactions between two continuous variables by dividing one into ten bins (deciles) or six bins (sextiles), with optional sample size display in legend for individual differences research.
Frequentist and Bayesian statistics are sometimes regarded as fundamentally different philosophies. Indeed, can both qualify as philosophies or is one of them just a pointless ritual? Is frequentist statistics only about $p$ values? Are frequentist estimates diametrically opposed to Bayesian posterior distributions? Are confidence intervals and credible intervals irreconcilable? Will R crash if lmerTest and brms are simultaneously loaded?
This post presents a run-through of a Bayesian workflow in R. The content is *closely* based on Bernabeu (2022), which was in turn based on lots of other references, also cited here.
The function knit_deleting_service_files() helps avoid (R) Markdown knitting errors caused by files and folders remaining from previous knittings (e.g., manuscript.tex, ZHJhZnQtYXBhLlJtZA==.Rmd, manuscript.synctex.gz). The only obligatory argument for this function is the name of a .Rmd or .md file. The optional argument is a path to a directory containing this file. The function first offers szeleting potential service files and folders in the directory. A confirmation is required in the console (see screenshot below). Next, the document is knitted. Last, the function offers deleting potential service files and folders again.
Three methods for balancing reproducibility and efficiency in R Markdown documents: embedding fast code directly, sourcing long but fast code from external scripts, and pre-running slow code while loading saved results.