Conditional permutation importance revisited

Debeer, Dries; Strobl, Carolin

doi:10.1186/s12859-020-03622-2

Cited by 104 publications

(156 citation statements)

References 40 publications

Supporting

Mentioning

154

Contrasting

Unclassified

Order By: Relevance

“…The Transitivity Index was calculated by training 1000 random forests with 3000 trees each with the party package [29]. The conditional variable importance measure [30] was then calculated for each of the 1000 random forests with the permimp package [31].…”

Section: The Transitivity Indexmentioning

confidence: 99%

The Transitivity Index: Using Transitivity as a continuous measure to account for clitic case alternation in Spanish causative constructions

Guajardo

2021

PLoS ONE

View full text Add to dashboard Cite

In Spanish causative constructions with dejar ‘let’ and hacer ‘make’ the subject of the embedded infinitive verb can appear in the accusative or the dative case. This case alternation has been accounted for by resorting to the notion of direct vs. indirect causation. Under this account, the accusative clitic with a transitive verb denotes direct causation while the dative clitic with an intransitive verb expresses indirect causation. The problem with this account is that we lack an independent definition of (in)direct causation in this context and so this approach suffers from circularity: the case of the clitic is used to determine causation type and causation type implies use of one or the other grammatical case. Therefore, a more objective way to account for clitic case alternation is needed. In this paper, I offer one possible solution in this direction by investigating clitic case alternation against Hopper and Thompson’s Transitivity parameters and a small number of other linguistic variables. The novelty of this approach is that I operationalise Transitivity as a weighted continuous measure (which I call the Transitivity Index) and use it to predict the case of the clitic. The results indicate that the transitivity of the infinitive verb, the animacy of the object and the agentivity of the subject are strong predictors of clitic case. Moreover, the Transitivity Index clearly shows that higher levels of Transitivity are associated with the dative clitic contrary to other contexts in which accusative is said to be more transitive. The findings in this paper allow us to arrive at a finer-grained characterization of the contexts in which each clitic case is more likely to occur and provide further evidence of the pervasiveness of Transitivity in natural language.

show abstract

Section: The Transitivity Indexmentioning

confidence: 99%

The Transitivity Index: Using Transitivity as a continuous measure to account for clitic case alternation in Spanish causative constructions

Guajardo

2021

PLoS ONE

View full text Add to dashboard Cite

show abstract

“…The relative importance (i.e. success), of each predictor was determined using the permimp [155] package. Fig 4 presents these values.…”

Section: Plos Onementioning

confidence: 99%

Variation isn’t that hard: Morphosyntactic choice does not predict production difficulty

et al. 2021

View full text Add to dashboard Cite

The following paper explores the link between production difficulty and grammatical variability. Using a sub-sample of the Switchboard Corpus of American English (285 transcripts, 34 speakers), this paper shows that the presence of variable contexts does not positively correlate with two metrics of production difficulty, namely filled pauses (um and uh) and unfilled pauses (speech planning time). When 20 morphosyntactic variables are considered collectively (N= 6,268), there is no positive effect. In other words, variable contexts do not correlate with measurable production difficulties. These results challenge the view that grammatical variability is somehow sub-optimal for speakers, with additional burdensome cognitive planning.

show abstract

“…A new implementation of this conditional permutation importance allows, in addition to faster execution, to vary the cursor between marginal and more partial measurements by means of a parameter (Debeer and Strobl 2020). The comparison of marginal vs more partial variable importances can be particularly useful from an interpretative point of view.…”

Section: Figure 4 Variable Importance Of a Random Forest Of The Us 2mentioning

confidence: 99%

“…Note: The computation of variable importances makes use of the conditional permutation scheme described in Debeer and Strobl (2020).…”

Section: Figure 9 Variable Importance For Surviving a First Filmmentioning

confidence: 99%

Trees and forests. Recursive partitioning as an alternative to parametric regression models in social sciences

Robette¹

2021

Preprint

View full text Add to dashboard Cite

Parametric regression models became the dominant tool of quantitative sociology. This dominance is not without its challenges and many criticisms have been expressed, both statistically and epistemologically. Still, the development of data mining, and then of machine learning, has led to the emergence of methodological approaches that make it possible to overcome most of the limitations of parametric regression models, for the various types of use that are of interest to the social sciences. We argue that recursive partitioning in particular may be highly vauable for social sciences. Indeed, this approach has a number of technical advantages over parametric regression and, above all, it is consistent with a conception of social determinations in terms of configurations of interdependent factors (and not of additions of independent factors). In a second step, we review a range of tools for interpreting the results obtained from recursive partitioning algorithms. Together, they form a very complete toolbox for the social sciences and show that recursive partitioning is no longer a black box as soon as the appropriate interpretative tools are mobilized. Finally, we illustrate the methods presented using sociological examples from the world of cinema. In doing so, we will show that these methods make it possible to deal with the different types of problems that arise in the social sciences when parametric regressions are usually used, in this case the study of structure effects and the ranking of explanatory factors.

show abstract

Conditional permutation importance revisited

Cited by 104 publications

References 40 publications

The Transitivity Index: Using Transitivity as a continuous measure to account for clitic case alternation in Spanish causative constructions

The Transitivity Index: Using Transitivity as a continuous measure to account for clitic case alternation in Spanish causative constructions

Variation isn’t that hard: Morphosyntactic choice does not predict production difficulty

Trees and forests. Recursive partitioning as an alternative to parametric regression models in social sciences

Contact Info

Product

Resources

About