Jan Schepers

Selecting Among Multi-Mode Partitioning Models of Different Complexities: A Comparison of Four Model Selection Criteria

Multi-mode partitioning models for N-way N-mode data reduce each of the N modes in the data to a small number of clusters that are mutually exclusive. Given a specific N-mode data set, one may wonder which multi-mode partitioning model (i.e., with which numbers of clusters for each mode) yields the most useful description of this data set and should therefore be selected. In this paper, we address this issue by investigating four possible model selection heuristics: multi-mode extensions of Calinski and Harabasz’s (1974) and Kaufman and Rousseeuw’s (1990) indices for one-mode k-means clustering and multi-mode partitioning versions of Timmerman and Kiers’s (2000) DIFFIT and Ceulemans and Kiers’s (2006) numerical convex hull based model selection heuristic for three-mode principal component analysis. The performance of these four heuristics is systematically compared in a simulation study, which shows that the DIFFIT and numerical convex hull heuristics perform satisfactory in the two-mode partitioning case and very good in the threemode partitioning case.

Three-mode partitioning

The three-mode partitioning model is a clustering model for three-way three-mode data sets that implies a simultaneous partitioning of all three modes involved in the data. In the associated data analysis, a data array is approximated by a model array that can be represented by a three-mode partitioning model of a prespecified rank, minimizing a least squares loss function in terms of differences between data and model. Algorithms have been proposed for this minimization, but their performance is not yet clear. A framework for alternating least-squares methods is described in order to offset the performance problem. Furthermore, a number of both existing and novel algorithms are discussed within this framework. An extensive simulation study is reported in which these algorithms are evaluated and compared according to sensitivity to local optima. The recovery of the truth underlying the data is investigated in order to assess the optimal estimates. The ordering of the algorithms with respect to performance in finding the optimal solution appears to change as compared to the results obtained from the simulation study when a collection of four empirical data sets have been used. This finding is attributed to violations of the implicit stochastic model underlying both the least-squares loss function and the simulation study. Support for the latter attribution is found in a second simulation study.

TwoMP: A MATLAB graphical user interface for two-mode partitioning

Two-way two-mode data occur in almost every domain of scientific psychology. The information present in such data, however, may be hard to grasp because of the dimensions of one or both modes. Two-mode partitioning addresses this problem by breaking down both modes into a number of mutually exclusive and exhaustive subsets. Although such a technique may be very useful, up to now, software—and consequently, two-mode partitioning—has been available only to a handful of specialists in the field. In this article, we present a free, easy-to-use MATLAB graphical user interface (TwoMP) for two-mode partitioning models, specifically developed for nonspecialist users. A short formal introduction is given on the statistics of the method, and the basic use of TwoMP is demonstrated with an example.

A unifying model involving a categorical and/or dimensional reduction for multimode data

Abstract A unifying model is presented that implies a categorical and/or dimensional reduction of one or several modes of a multiway data set. The model encompasses a broad range of (existing as well as to be developed) discrete, continuous, as well as hybrid discrete–continuous reduction models as special cases, which all imply a decomposition of the reconstructed data on the basis of quantifications of the different data modes and a linking array. An analysis of the objective or loss function associated with the model leads to two generic algorithmic strategies, the possibilities and limitations of which are the object of a subsequent discussion.

Recovery 3 and 12 months after hysterectomy: epidemiology and predictors of chronic pain, physical functioning, and global surgical recovery

AbstractChronic postsurgical pain (CPSP) is 1 important aspect of surgical recovery. To improve perioperative care and postoperative recovery knowledge on predictors of impaired recovery is essential. The aim of this study is to assess predictors and epidemiological data of CPSP, physical functioning (SF-36PF, 0–100), and global surgical recovery (global surgical recovery index, 0–100%) 3 and 12 months after hysterectomy for benign indication.A prospective multicenter cohort study was performed. Sociodemographic, somatic, and psychosocial data were assessed in the week before surgery, postoperatively up to day 4, and at 3- and 12-month follow-up. Generalized linear model (CPSP) and linear-mixed model analyses (SF-36PF and global surgical recovery index) were used. Baseline data of 468 patients were collected, 412 (88%) patients provided data for 3-month evaluation and 376 (80%) patients for 12-month evaluation.After 3 and 12 months, prevalence of CPSP (numeric rating scale ≥ 4, scale 0–10) was 10.2% and 9.0%, respectively, SF-36PF means (SD) were 83.5 (20.0) and 85.9 (20.2), global surgical recovery index 88.1% (15.6) and 93.3% (13.4). Neuropathic pain was reported by 20 (5.0%) patients at 3 months and 14 (3.9%) patients at 12 months. Preoperative pain, surgery-related worries, acute postsurgical pain on day 4, and surgery-related infection were significant predictors of CPSP. Baseline level, participating center, general psychological robustness, indication, acute postsurgical pain, and surgery-related infection were significant predictors of SF-36PF. Predictors of global surgical recovery were baseline expectations, surgery-related worries, American Society of Anesthesiologists classification, type of anesthesia, acute postsurgical pain, and surgery-related infection.Several predictors were identified for CPSP, physical functioning, and global surgical recovery. Some of the identified factors are modifiable and optimization of patients’ preoperative pain status and psychological condition as well as reduction of acute postsurgical pain and surgery-related infection may lead to improvement of outcome.

A Two-Mode Clustering Method to Capture the Nature of the Dominant Interaction Pattern in Large Profile Data Matrices

Profile data abound in a broad range of research settings. Often it is of considerable theoretical importance to address specific structural questions with regard to the major pattern as included in such data. A key challenge in this regard pertains to identifying which type of interaction (double ordinal, mixed ordinal/disordinal, double disordinal) most adequately fits the major pattern in a profile data set at hand. In the present article a novel methodology is proposed to deal with this challenge. This methodology is based on constrained and unconstrained versions of a recently introduced 2-mode clustering model, the real-valued hierarchical classes model. The methodology is illustrated using empirical Person × Situation profile data on altruism.

Stability of and Associations Between Social-Cognitive Determinants Over Time

This study reports on the stability of social-cognitive determinants, and on associations between social-cognitive determinants to show insight in the theory of planned behavior (TPB). In all, 278 health professionals who encourage patients to become physically active completed online TPB-based surveys at baseline (Time 1 [T1]) and six months later (Time 2 [T2]). No intervention took place. No differences were found for all social-cognitive determinants measured at T1 compared with T2 (6 months later), except for intention (t test = 5.18, p < .001). Structural equation modeling—χ2(5, N = 278) = 2.35, p = .80, root mean square error of approximation = 0.00—showed that behavior T1 and attitude T1 predicted intention T1 (R2 = .57, p = <.001); that behavior T1 and barriers T1 predicted behavior T2 (R2 = .38, p = <.001); and that behavior T2, intention T1, and attitude T1 predicted intention T2 (R2 =.60, p = <.001). Intention T1 did not predict behavior T2. The model achieved a good fit with the data. Findings revealed that social-cognitive determinants remained stable over time, with intention being instable. Without intervention, the intention decreased, while the social-cognitive determinants (attitudes, perceived behavioral control, and subjective norms) for intention and the corresponding behavior remained unchanged. For intervention development it seems important to value health professionals’ previous or past encouraging behavior (T1), this to change intention and behavior, or to initiate new behavior. Behavior T1 showed a predictive variable and predicted attitude T1, intention T1, barriers T1, and behavior T2. Barriers that obstruct health professionals’ encouraging behavior are encountered, and barriers influence attitudes T1 and the behavior T2 to encourage patients.

Supporting Health Care Professionals to Improve the Processes of Shared Decision Making and Self-Management in a Web-Based Intervention: Randomized Controlled Trial

Background Research to assess the effect of interventions to improve the processes of shared decision making and self-management directed at health care professionals is limited. Using the protocol of Intervention Mapping, a Web-based intervention directed at health care professionals was developed to complement and optimize health services in patient-centered care. Objective The objective of the Web-based intervention was to increase health care professionals’ intention and encouraging behavior toward patient self-management, following cardiovascular risk management guidelines. Methods A randomized controlled trial was used to assess the effect of a theory-based intervention, using a pre-test and post-test design. The intervention website consisted of a module to help improve professionals’ behavior, a module to increase patients’ intention and risk-reduction behavior toward cardiovascular risk, and a parallel module with a support system for the health care professionals. Health care professionals (n=69) were recruited online and randomly allocated to the intervention group (n=26) or (waiting list) control group (n=43), and invited their patients to participate. The outcome was improved professional behavior toward health education, and was self-assessed through questionnaires based on the Theory of Planned Behavior. Social-cognitive determinants, intention and behavior were measured pre-intervention and at 1-year follow-up. Results The module to improve professionals’ behavior was used by 45% (19/42) of the health care professionals in the intervention group. The module to support the health professional in encouraging behavior toward patients was used by 48% (20/42). The module to improve patients’ risk-reduction behavior was provided to 44% (24/54) of patients. In 1 of every 5 patients, the guideline for cardiovascular risk management was used. The Web-based intervention was poorly used. In the intervention group, no differences in social-cognitive determinants, intention and behavior were found for health care professionals, compared with the control group. We narrowed the intervention group and no significant differences were found in intention and behavior, except for barriers. Results showed a significant overall difference in barriers between the intervention and the control group (F 1=4.128, P=.02). Conclusions The intervention was used by less than half of the participants and did not improve health care professionals’ and patients’ cardiovascular risk-reduction behavior. The website was not used intensively because of time and organizational constraints. Professionals in the intervention group experienced higher levels of barriers to encouraging patients, than professionals in the control group. No improvements were detected in the processes of shared decision making and patient self-management. Although participant education level was relatively high and the intervention was pre-tested, it is possible that the way the information was presented could be the reason for low participation and high dropout. Further research embedded in professionals’ regular consultations with patients is required with specific emphasis on the processes of dissemination and implementation of innovations in patient-centered care. Trial Registration Netherlands Trial Register Number (NTR): NTR2584; http://www.trialregister.nl/trialreg/admin/rctview.asp?TC=2584 (Archived by WebCite at http://www.webcitation.org/6STirC66r).

The Real-Valued Model of Hierarchical Classes

We propose a non-negative real-valued model of hierarchical classes (HICLAS) for two-way two-mode data. Like the other members of the HICLAS family, the non-negative real-valued model (NNRV-HICLAS) implies simultaneous hierarchically organized classifications of all modes involved in the data. A distinctive feature of the novel model is that it yields continuous, non-negative real-valued reconstructed data, which considerably expands the application range of the HICLAS family. The expansion implies a major algorithmic challenge as it involves a move from the typical discrete optimization problems in HICLAS to a mixed discrete-continuous one. To solve this mixed discrete-continuous optimization problem, a two-stage algorithm combining a simulated annealing and an alternating local descent stage is proposed. Subsequently it is evaluated in a simulation study. Finally, the NNRVHICLAS model is applied to an empirical data set on anger.

A unifying model for biclustering

A unifying biclustering model is presented for the simultaneous classification of the rows and columns of a rectangular data matrix. The model encompasses a broad range of (existing as well as to be developed) biclustering models as special cases, which all imply homogeneous data clusters on the basis of which the data can be reconstructed making use of a Sum- or Max-operator. An analysis of the objective or loss function associated with the model leads to two generic algorithmic strategies. In the discussion, we point at various possible model extensions.