Florian Wickelmaier

A Matlab function to estimate choice model parameters from paired-comparison data

Tversky (1972) has proposed a family of models for paired-comparison data that generalize the Bradley—Terry—Luce (BTL) model and can, therefore, apply to a diversity of situations in which the BTL model is doomed to fail. In this article, we present a Matlab function that makes it easy to specify any of these general models (EBA, Pretree, or BTL) and to estimate their parameters. The program eliminates the time-consuming task of constructing the likelihood function by hand for every single model. The usage of the program is illustratedby several examples. Features of the algorithm are outlined. The purpose of this article is to facilitate the use of probabilistic choice models in the analysis of data resulting from paired comparisons.

Short-term treatment with risperidone or haloperidol in first-episode schizophrenia: 8-week results of a randomized controlled trial within the German Research Network on Schizophrenia

Patients with first-episode schizophrenia appear to respond to lower doses of neuroleptics, and to be more sensitive to developing extrapyramidal side-effects. The authors therefore compared in such patients the efficacy and extrapyramidal tolerability of comparatively low dosages of the atypical neuroleptic risperidone and of the conventional neuroleptic haloperidol. Risperidone was hypothesized to have better extrapyramidal tolerability and efficacy in treating negative symptoms. Patients were randomly assigned under double-blind conditions to receive risperidone (n=143) or haloperidol (n=146) for 8 wk. The primary efficacy criterion was the estimated difference in the mean change in the Positive and Negative Symptom Scale (PANSS) negative score between treatment groups; secondary efficacy criteria were changes on the PANSS total score and other PANSS subscores, and several other measures of psychopathology and general functioning. The primary tolerability criterion was the difference in baseline-adjusted occurrence rates of extrapyramidal side-effects measured with the Simpson-Angus Scale (SAS) compared between treatment groups. The main hypothesis was that risperidone would be superior in terms of improving negative symptoms and lowering the risk of extrapyramidal symptoms. Secondary tolerability criteria were the other extrapyramidal symptoms, measured with the Hillside Akathisia Scale (HAS) and the Abnormal Involuntary Movement Scale (AIMS). The average mean daily doses were 3.8 mg (s.d.=1.5) for risperidone and 3.7 mg (s.d.=1.5) for haloperidol. There were similar, significant improvements in both treatment groups in the primary and secondary efficacy criteria. At week 8 nearly all scores of extrapyramidal side-effects indicated a significantly higher prevalence of extrapyramidal side-effects with haloperidol than with risperidone [SAS: risperidone 36.5% of patients; haloperidol 51.5% of patients; likelihood ratio test, chi2(1)=7.8, p=0.005]. There were significantly fewer drop-outs [risperidone n=55, drop-out rate=38.5%; haloperidol n=79, drop-out rate=54.1%, chi2(1)=7.1, p=0.009] and a longer non-discontinuation time [risperidone: average of 50.8 d to drop-out; haloperidol: average of 44.0 d to drop-out; log rank test, chi2(1)=6.4, p=0.011] in the risperidone group. Risperidone and haloperidol appear to be equally effective in treating negative and other symptoms of first-episode schizophrenia. Risperidone has better extrapyramidal tolerability and treatment retention rate than the equivalent dose of haloperidol in these patients.

Evaluation of multichannel reproduced sound: scaling auditory attributes underlying listener preference.

A study was conducted with the goal of quantifying auditory attributes that underlie listener preference for multichannel reproduced sound. Short musical excerpts were presented in mono, stereo, and several multichannel formats to a panel of 40 selected listeners. Scaling of auditory attributes, as well as overall preference, was based on consistency tests of binary paired-comparison judgments and on modeling the choice frequencies using probabilistic choice models. As a result, the preferences of nonexpert listeners could be measured reliably at a ratio scale level. Principal components derived from the quantified attributes predict overall preference well. The findings allow for some generalizations within musical program genres regarding the perception of and preference for certain spatial reproduction modes, but for limited generalizations across selections from different musical genres.

Accounting for individual differences in Bradley-Terry models by means of recursive partitioning

The preference scaling of a group of subjects may not be homogeneous, but different groups of subjects with certain characteristics may show different preference scalings, each of which can be derived from paired comparisons by means of the Bradley-Terry model. Usually, either different models are fit in predefined subsets of the sample or the effects of subject covariates are explicitly specified in a parametric model. In both cases, categorical covariates can be employed directly to distinguish between the different groups, while numeric covariates are typically discretized prior to modeling. Here, a semiparametric approach for recursive partitioning of Bradley-Terry models is introduced as a means for identifying groups of subjects with homogeneous preference scalings in a data-driven way. In this approach, the covariates that—in main effects or interactions—distinguish between groups of subjects with different preference orderings, are detected automatically from the set of candidate covariates. One main advantage of this approach is that sensible partitions in numeric covariates are also detected automatically.

Atypical symptoms in hospitalised patients with major depressive episode: frequency, clinical characteristics, and internal validity

OBJECTIVE The objective was (1) to assess the frequency of atypical depression (AD) in depressed inpatients; (2) to compare clinical features of patients with atypical and nonatypical depression (Non-AD) (3) to evaluate the meaning of single psychopathological symptoms with special respect to mood reactivity. METHOD Diagnoses of 1073 inpatients were assessed according to DSM-IV using SCID (Structured Clinical Interview for the DSM-IV) and AMDP (Association for Methodology and Documentation). Diagnosis of atypical depression was defined according to criteria of the DSM-IV specifier for AD. All patients were rated using HAMD-21 (Hamilton Depression Scale). RESULTS A high percentage of patients met criteria for AD (15.3%, 95% CI 13.0-17.9%). Women were more likely to suffer from AD (OR=1.54, p=0.037). There were no significant differences between AD and Non-AD patients regarding age, HAMD total baseline score, and diagnosis of any bipolar illness. In terms of psychopathology patients with AD were significantly more likely to suffer from somatic anxiety, somatic symptoms, guilt, genital symptoms, depersonalisation and suspiciousness as defined by HAMD-21 items. Interestingly, mood reactivity was not found to be significantly associated with the presence of two or more additional symptoms of AD. LIMITATIONS Results were assessed by a post-hoc analysis, based on prospectively collected data. Compared to other inpatient samples with MDE, prevalence of bipolar disorder was rather low. CONCLUSION (1) Frequency of AD may be underestimated, especially in inpatient samples. Further studies of inpatient samples are recommended. (2) Quality of distinct anxiety symptoms may be different in both groups, with AD patients being more likely to suffer from somatic symptoms and somatic anxiety. The presence of suspiciousness and even paranoid phenomena may not exclude a diagnosis of AD, but may be related to rejection sensitivity. (3) The mandatory presence of mood reactivity for the diagnosis of AD needs further consideration, regarding its validity for the concept.

Minimum Discrepancy Estimation in Probabilistic Knowledge Structures

Practical applications of the theory of knowledge structures often rely on a probabilistic version, known as the basic local independence model. The paper outlines various procedures for estimating its parameters, including maximum likelihood (ML) via the expectation-maximization (EM) algorithm, the computationally efficient minimum discrepancy (MD) estimation as well as MDML, a hybrid method combining the two approaches.

Deriving auditory features from triadic comparisons

A feature-based representation of auditory stimuli is proposed and tested experimentally. Within a measurement-theoretical framework, it is possible to decide whether a representation of subjective judgments with a set of auditory features is possible and how unique such a representation is. Furthermore, the method avoids confounding listeners’ perceptual and verbal abilities, in that it strictly separates the process of identifying auditory features from labeling them. The approach was applied to simple synthetic sounds with well-defined physical properties (narrow-band noises and complex tones). For each stimulus triad, listeners had to judge whether the first two sounds displayed a common feature that was not shared by the third, by responding with a simple “yes” or “no.” Because of the high degree of consistency in the responses, feature structures could be obtained for most of the subjects. In summary, the proposed procedure constitutes a supplement to the arsenal of psychometric methods with the main focus of identifying the type of sensation itself, rather than of measuring its threshold or magnitude.

Scaling the perceived fluctuation strength of frequency‐modulated tones

Fluctuation strength is one of the major psychoacoustic variables considered in sound‐quality evaluation. Zwicker and Fastl [Psychoacoustics (Springer, Berlin, 1999)] summarize recommendations for its computation, which have already been implemented in various software applications, even though the data basis is rather limited. In particular, the dependency of fluctuation strength on modulation frequency and modulation depth has seemingly never been tested in a factorial design. Therefore, in experiment I both of these factors were varied simultaneously in order to create 54 different frequency‐modulated sinusoids. The task of the subjects was to directly estimate the perceived magnitude of fluctuation strength. The results do not conform well with the prevalent model of fluctuation strength. In experiment II this finding was further investigated by varying only one factor at a time. The results show that large individual differences, particularly in the effect of modulation frequency, persist. Thus, in e...

Using Recursive Partitioning to Account for Parameter Heterogeneity in Multinomial Processing Tree Models

In multinomial processing tree (MPT) models, individual differences between the participants in a study can lead to heterogeneity of the model parameters. While subject covariates may explain these differences, it is often unknown in advance how the parameters depend on the available covariates, that is, which variables play a role at all, interact, or have a nonlinear influence, etc. Therefore, a new approach for capturing parameter heterogeneity in MPT models is proposed based on the machine learning method MOB for model-based recursive partitioning. This procedure recursively partitions the covariate space, leading to an MPT tree with subgroups that are directly interpretable in terms of effects and interactions of the covariates. The pros and cons of MPT trees as a means of analyzing the effects of covariates in MPT model parameters are discussed based on simulation experiments as well as on two empirical applications from memory research. Software that implements MPT trees is provided via the mpttree function in the psychotree package in R.

Construction and Updating of Event Models in Auditory Event Processing.

Humans segment the continuous stream of sensory information into distinct events at points of change. Between 2 events, humans perceive an event boundary. Present theories propose changes in the sensory information to trigger updating processes of the present event model. Increased encoding effort finally leads to a memory benefit at event boundaries. Evidence from reading time studies (increased reading times with increasing amount of change) suggest that updating of event models is incremental. We present results from 5 experiments that studied event processing (including memory formation processes and reading times) using an audio drama as well as a transcript thereof as stimulus material. Experiments 1a and 1b replicated the event boundary advantage effect for memory. In contrast to recent evidence from studies using visual stimulus material, Experiments 2a and 2b found no support for incremental updating with normally sighted and blind participants for recognition memory. In Experiment 3, we replicated Experiment 2a using a written transcript of the audio drama as stimulus material, allowing us to disentangle encoding and retrieval processes. Our results indicate incremental updating processes at encoding (as measured with reading times). At the same time, we again found recognition performance to be unaffected by the amount of change. We discuss these findings in light of current event cognition theories.