I am dealing with nested data, and I remember from an article by Clark (1973) that nested should be analysed using special models. I’ve looked into mixed-effects models, and I’ve reached a structure with random intercepts by subjects and by items. Is this fine?
In early days, researchers would aggregate the data across these repeated measures to prevent the violation of the assumption of independence of observations, which is one of the most important assumptions in statistics.
Notwithstanding the need for speed in most scientific projects, what should be the minimum acceptable standards in the creation of research materials such as stimuli, custom software for data collection (e.g., experiment in jsPsych, OpenSesame or psychoPy), or scripts for statistical analysis?
The present script can be used to pre-process data from a frequency list of the Norwegian as Web Corpus (NoWaC).
Before using the script, the frequency list should be downloaded from https://www.hf.uio.no/iln/english/about/organization/text-laboratory/services/nowac-frequency.html. The list is described as ‘frequency list sorted primary alphabetic and secondary by frequency within each character’, and the direct URL is: https://www.tekstlab.uio.no/nowac/download/nowac-1.1.lemma.frek.sort_alf_frek.txt.gz. The download requires signing in to an institutional network. Last, the downloaded file should be unzipped.
The R package ‘simr’ has greatly facilitated power analysis for mixed-effects models using Monte Carlo simulation (which involves running hundreds or thousands of tests under slight variations of the data). The powerCurve function is used to estimate the statistical power for various sample sizes in one go. Since the tests are run serially, they can take a VERY long time; approximately, the time it takes to run the model supplied once (say, a few hours) times the number of simulations (nsim, which should be higher than 200), and times the number of different sample sizes examined.
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 two continuous variables.
#> Loading required package: Matrix
#> Learn more about sjPlot with 'browseVignettes("sjPlot")'.
# Create data partially based on code by Ben Bolker # from https://stackoverflow.
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.
#> Loading required package: Matrix
#> Install package "strengejacke" from GitHub (`devtools::install_github("strengejacke/strengejacke")`) to load all sj-packages at once!
To assess whether convergence warnings render the results invalid, or on the contrary, the results can be deemed valid in spite of the warnings, Bates et al. (2023) suggest refitting models affected by convergence warnings with a variety of optimizers. The authors argue that, if the different optimizers produce practically-equivalent results, the results are valid. The allFit function from the ‘lme4’ package allows the refitting of models using a number of optimizers.
Here’s an example of fuzzy-matching strings in R that I shared on StackOverflow. In stringdist_join, the max_dist argument is used to constrain the degree of fuzziness.
#> #> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> #> filter, lag
#> The following objects are masked from 'package:base':
#> #> intersect, setdiff, setequal, union
small_tab = data.frame(Food.Name = c('Corn', 'Squash', 'Peppers'), Food.Code = c(NA, NA, NA))
large_tab = data.
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.
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. Two of these functions are deciles_interaction_plot and sextiles_interaction_plot. These functions allow the plotting of interactions between two continuous variables.
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.
As technology and research methods advance, the data sets tend to be larger and the methods more exhaustive. Consequently, the analyses take longer to run. This poses a challenge when the results are to be presented using R Markdown. One has to balance reproducibility and efficiency. On the one hand, it is desirable to keep the R Markdown document as self-contained as possible, so that those who may later examine the document can easily test and edit the code.
When knitting an R Markdown document after the first time, errors may sometimes appear. Three tips are recommended below.
1. Close PDF reader window
When the document is knitted through the ‘Knit’ button, a PDF reader window opens to present the result. Closing this window can help resolve errors.
2. Delete service files
Every time the Rmd is knitted, some service files are created. Some of these files have the ‘.
The powercurve function from the R package ‘simr’ (Green & MacLeod, 2016) can incur very long running times when the method used for the calculation of p values is Kenward-Roger or Satterthwaite (see Luke, 2017). Here I suggest three ways for cutting down this time.
Where possible, use a high-performance (or high-end) computing cluster. This removes the need to use personal computers for these long jobs.
In case you’re using the fixed() parameter of the powercurve function, and calculating the power for different effects, run these at the same time (‘in parallel’) on different machines, rather than one after another.
This open-source, R-based web application allows the conversion of video captions (subtitles) from the Web Video Text Tracks (WebVTT) Format into plain texts. For this purpose, users upload a WebVTT file with the extension of 'vtt' or 'txt'. …
This document is part of teaching materials created for the workshop 'Open data and reproducibility v2.1: R Markdown, dashboards and Binder', delivered at the CarpentryCon 2020 conference. The purpose of this specific document is to practise R Markdown, including basic features such as Markdown markup and code chunks, along with more special features such as cross-references for figures, tables, code chunks, etc. Since this conference was originally going to take place in Madison, let's look at some open data from the City of Madison.
This open-source, R-based web application is suitable for educational or research purposes in experimental sciences. It allows the **creation of varied data sets with specified structures, such as between-group or within-participant variables, that …
Dashboard with open data from a study by Prudic et al. (2018), that compares citizen science with traditional methods in butterfly sampling. Coding tasks included long-transforming, merging, and as ever, wrangling with a table.
This project offers free activities to learn and practise reproducible data presentation. Pablo Bernabeu organises these events in the context of a Software Sustainability Institute Fellowship. Programming languages such as R and Python offer free, powerful resources for data processing, visualisation and analysis. Experience in these programs is highly valued in data-intensive disciplines. Original data has become a public good in many research fields thanks to cultural and technological advances. On the internet, we can find innumerable data sets from sources such as scientific journals and repositories (e.g., OSF), local and national governments, non-governmental organisations (e.g., data.world), etc. Activities comprise free workshops and datathons.
Principal Component Analysis (PCA) is a technique used to find the core components that underlie different variables. It comes in very useful whenever doubts arise about the true origin of three or more variables. There are two main methods for performing a PCA: naive or less naive. In the naive method, you first check some conditions in your data which will determine the essentials of the analysis. In the less-naive method, you set those yourself based on whatever prior information or purposes you had. The 'naive' approach is characterized by a first stage that checks whether the PCA should actually be performed with your current variables, or if some should be removed. The variables that are accepted are taken to a second stage which identifies the number of principal components that seem to underlie your set of variables.
This app presents linguistic data over several tabs. The code combines the great front-end of Flexdashboard—based on R Markdown and yielding an unmatched user interface—, with the great back-end of Shiny—allowing users to download sections of data they select, in various formats. The hardest nuts to crack included modifying the rows/columns orientation without affecting the functionality of tables. A cool, recent finding was the reactable package. A nice feature, allowed by Flexdashboard, was the use of quite different formats in different tabs.
Dashboards for data visualisation, such as R Shiny and Tableau, allow the interactive exploration of data by means of drop-down lists and checkboxes, with no coding required from the final users. These web applications run on internet browsers, allowing for three viewing modes, catered to both analysts and the public at large: (1) private viewing (useful during analysis), (2) selective sharing (used within work groups), and (3) internet publication. Among the available platforms, R Shiny and Tableau stand out due to being relatively accessible to new users. Apps serve a broad variety of purposes. In science and beyond, these apps allow us to go the extra mile in sharing data. Alongside files and code shared in repositories, we can present the data in a website, in the form of plots or tables. This facilitates the public exploration of each section of the data (groups, participants, trials...) to anyone interested, and allows researchers to account for their proceeding in the analysis.
We tested whether conceptual processing is modality-specific by tracking the time course of the Conceptual Modality Switch effect. Forty-six participants verified the relation between property words and concept words. The conceptual modality of consecutive trials was manipulated in order to produce an Auditory-to-visual switch condition, a Haptic-to-visual switch condition, and a Visual-to-visual, no-switch condition. Event-Related Potentials (ERPs) were time-locked to the onset of the first word (property) in the target trials so as to measure the effect online and to avoid a within-trial confound. A switch effect was found, characterized by more negative ERP amplitudes for modality switches than no-switches. It proved significant in four typical time windows from 160 to 750 milliseconds post word onset, with greater strength in the Slow group, in posterior brain regions, and in the N400 window. The earliest switch effect was located in the language brain region, whereas later it was more prominent in the visual region. In the N400 and Late Positive windows, the Quick group presented the effect especially in the language region, whereas the Slow had it rather in the visual region. These results suggest that contextual factors such as time resources modulate the engagement of linguistic and embodied systems in conceptual processing.