Statistics for linguist(ic)s blog

Exporting R graphics: A basic workflow

data visualization

In this blog post, I describe my workflow for exporting and polishing graphs drawn in R.

Modeling clustered frequency data II: Texts of disproportionate length

corpus linguistics

regression

clustered data

frequency data

bias

imbalance

negative binomial

This blog post illustrates a number of strategies for modeling clustered count data. It describes how they handle the non-independence among observations and what kind of estimates they return. The focus is on a situation where texts have very different lengths.

Modeling clustered frequency data I: Texts of similar length

corpus linguistics

regression

clustered data

frequency data

negative binomial

This blog post illustrates a number of strategies for modeling clustered count data. It describes how they handle the non-independence among observations and what kind of estimates they return. The focus is on a situation where texts have roughly the same length.

Frequency estimates based on random-intercept Poisson models

corpus linguistics

regression

clustered data

frequency data

negative binomial

Clustered count data can be modeled using a Poisson regression model including random intercepts. This blog post describes how this model represents the data and the different kinds of frequency estimates it produces.

Modeling clustered binomial data

corpus linguistics

regression

clustered data

binary data

This blog post illustrates a number of strategies for modeling clustered binomial data. It describes how they handle the non-independence among observations and what kind of estimates they return.

Imbalance across predictor levels affects data summaries

Obstacles to replication in corpus linguistics

corpus linguistics

replication crisis

regression

bias

imbalance

This blog post is part of a small series on obstacles to replication in corpus linguistics. It deals with problems that can arise if the observations drawn from a corpus are unbalanced across relevant subgroups in the data. I show how simple and comparative data summaries can vary depending on whether we (unintentionally) calculate weighted averages, or adjust estimates for imbalances by taking a simple average across subgroups. As these are two different estimands, the choice affects the comparability of studies – including an original study and its direct replication.

Clustering in the data affects statistical uncertainty intervals

Obstacles to replication in corpus linguistics

corpus linguistics

replication crisis

regression

clustered data

This blog post is part of a small series on obstacles to replication in corpus linguistics. It deals with a prevalent issue in corpus data analysis: the non-independence of data points that results from clustered (or hierarchical) data layouts. I show how an inadequate analysis can produce unduly narrow expectations of a replication study.

Unbalanced distributions and their consequences: Speakers in the Spoken BNC2014

corpus linguistics

clustered data

negative binomial

clustered data

imbalance

This blog post illustrates how the disproportionate representation of speakers in a corpus can lead to distorted results if the source of data points (i.e. the speaker ID) is not taken into account in the analysis.

Modeling the interpretation of quantifiers using beta regression

regression

distributional modeling

This blog post shows how to use beta regression to model the proportional interpretation of the quantifiers few, some, many, and most. We consider variable-dispersion and mixed-effects structures as well as diagnostics for frequentist and Bayesian models.

Different parameterizations of the negative binomial distribution

corpus linguistics

dispersion

negative binomial

This blog post discusses two different parameterizations of the negative binomial distribution and groups R packages (and functions) based on the version they implement.

The negative binomial distribution: A visual explanation

corpus linguistics

dispersion

negative binomial

This blog post uses a visual approach to explain how the negative binomial distribution works.

A computational shortcut for the dispersion measure D_A

corpus linguistics

dispersion

This short blog post draws attention to the computational shortcut given in Wilcox (1973) for calculating the dispersion measure D_A.

The replication crisis: Implications for myself

replication crisis

open science

In this blog post, I reflect on the ways in which learning about the replication crisis in science has affected my own work.

Structured down-sampling: Implementation in R

corpus linguistics

down-sampling

This blog post shows how to implement structured down-sampling in R.

Two types of down-sampling in corpus-based work

corpus linguistics

down-sampling

This short blog post contrasts the different ways in which the term down-sampling is used in corpus-based work.

‘Dispersion’ in corpus linguistics and statistics

corpus linguistics

dispersion

This blog post clarifies the different ways in which the term dispersion is used in corpus linguistics and statistics.