A comprehensive framework of regression models for ordinal data

A Novel Indicator to Correct for Individual Reported Heterogeneity. An Application to Self-Evaluation of Later-Life Depression

Program evaluations often investigate complex or multi-dimensional constructs, such as individual opinions or attitudes, by means of ratings. A different interpretation of the same question may affect cross-country comparability, leading to the Differential Item Functioning problem. Anchoring vignettes were introduced in the literature as a way to adjust self-evaluations from this interpersonal incomparability. In this paper, we first introduce a new nonparametric solution to analyse anchoring vignette data, recoding a variable based on a rating scale to a new corrected-variable that guarantees comparability in any cross-country analysis. Then, we exploit the flexibility of a mixture model introduced to account for uncertainty in the response process (the CUP model) to test if the proposed solution is effectively able to remove this reported heterogeneity. This solution is easy to construct and has important advantages compared with the original nonparametric solution adopted with anchoring vignette data. The novel indicator is applied to investigate self-reported depression in an old population. Data that will be analysed come from the second wave of the Survey of Health, Ageing and Retirement in Europe, collected in 2006/2007. Results highlight the need of correcting for reported heterogeneity comparing individual self-evaluations. Once interpersonal incomparability resulting from the different uses of response scales is removed from the self-assessments, some estimates are reversed in magnitude and signs with respect to the analysis of the collected data.

Consumer perception of attributes of organic food in Italy: A CUB model study

Organic food, consumers and their buying behaviour are well examined fields of research, although there is a lack of consistent findings on consumers' perception about organic food's quality, in terms of healthiness, safety, and environmental sustainability, and on determinants of perceived quality. This study investigates how consumers perceive the quality of organic food, in terms of environmental sustainability, safety, and healthiness. The study also analyses how and to what extent perceived quality of organic food is influenced by the presence of information related to quality on food products' labels and consumers' socio-demographic profile. A survey has been conducted on a convenience sample of Italian consumers, recruited through a snowball sampling technique. An approach based on a Combination of Uniform and shifted Binomial random variables, named CUB model, is adopted to analyse consumers' perceptions in terms of two latent components, feeling and uncertainty. The CUB model approach is suitable for analyses that involve consumers perception. The results suggest that consumers perceive safety of organic food better than healthiness and environmentally sustainable attributes. Findings also highlight that the presence of specific information on food's label contributes to perceive organic food as healthier, safe, and environmentally sustainable: the more the details on food labels, the higher the consumers' perception. Furthermore, consumers' socio-demographic profile plays a significant role: males and females have a different perception of organic food and younger consumers tend to be more prone to buy and consume organic product.

•

Consumers are more confident with healthiness and sustainability of organic food.

•

Males and females have a different perception of organic food.

•

Food labels increase the perception of organic food as healthy, safe, sustainable.

•

More details on labels of organic food enhance consumers' perception.

Uncertainty in Latent Trait Models

A model that extends the Rasch model and the Partial Credit Model to account for subject-specific uncertainty when responding to items is proposed. It is demonstrated that ignoring the subject-specific uncertainty may yield biased estimates of model parameters. In the extended version of the model, uncertainty and the underlying trait are linked to explanatory variables. The parameterization allows to identify subgroups that differ in uncertainty and the underlying trait. The modeling approach is illustrated using data on the confidence of citizens in public institutions.

Application of Beta-Distribution and Combined Uniform and Binomial Methods in Longitudinal Modeling of Bounded Outcome Score Data

Disease status is often measured with bounded outcome scores (BOS) which takes a discrete set of values on a finite range. The distribution of such data is often skewed, rendering the standard analysis methods assuming normal distribution inappropriate. Among the methods used for BOS analyses, two of them have the ability to predict the data within its natural range and accommodate data skewness: (1) a recently proposed beta-distribution based approach and (2) a mixture model known as CUB (combined uniform and binomial). This manuscript compares the two approaches, using an established mechanism-based longitudinal exposure-response model to analyze data from a phase 2 clinical trial in psoriatic patients. The beta-distribution-based approach was confirmed to perform well, and CUB also showed potential. A separate issue of modeling clinical trial data is that the collected baseline disease score range may be more limited than that of post-treatment disease score due to clinical trial inclusion criteria, a fact that is typically ignored in longitudinal modeling. The effect of baseline disease status restriction should in principle be adjusted for in longitudinal modeling.

Facilitating Longitudinal Exposure-Response Modeling of a Composite Endpoint Using the Joint Modeling of Sparsely and Frequently Collected Subcomponents

Longitudinal exposure-response modeling plays an important role in optimizing dose and dosing regimens in clinical drug development. Certain clinical trials contain induction and maintenance phases where the maintenance treatment depends on the subjects' achieving the main endpoint outcome in the induction phase. Due to logistic difficulties and cost considerations, the main endpoint is usually collected more sparsely than a subcomponent (or other related endpoints). The sparse collection of the main endpoint hampers its longitudinal modeling. In principle, the frequent collection of a subcomponent allows its longitudinal modeling. However, the model evaluation via the visual predictive check (VPC) in the maintenance phase is difficult due to the requirement of the main-endpoint model to identify the treatment subgroups. This manuscript proposes a solution to this dilemma via the joint modeling of the main endpoint and the subcomponent. The challenges are illustrated by analyzing the data collected up to 60 weeks from a phase III trial of ustekinumab in patients with moderate-to-severe ulcerative colitis (UC). The main endpoint Mayo score, a commonly used composite endpoint to measure the severity of UC, was collected only at baseline, the end of the induction phase, and the end of the maintenance phase. The partial Mayo score, which is a major subset of the Mayo score, was collected at nearly every 4 weeks. A longitudinal joint exposure-response model, developed under a latent-variable Indirect Response modeling framework, described the Mayo score time course and facilitated the VPC model evaluation under a response-adaptive trial design.

Modeling near-continuous clinical endpoint as categorical: application to longitudinal exposure-response modeling of Mayo scores for golimumab in patients with ulcerative colitis

Accurate characterization of exposure-response relationship of clinical endpoints is important in drug development to identify optimal dose regimens. Endpoints with ≥ 10 ordered categories are typically analyzed as continuous. This manuscript aims to show circumstances where it is advantageous to analyze such data as ordered categorical. The results of continuous and categorical analyses are compared in a latent-variable based Indirect Response modeling framework for the longitudinal modeling of Mayo scores, ranging from 0 to 12, which is commonly used as a composite endpoint to measure the severity of ulcerative colitis (UC). Exposure response modeling of Mayo scores is complicated by the fact that studies typically include induction and maintenance phases with re-randomizations and other response-driven dose adjustments. The challenges are illustrated in this work by analyzing data collected from 3 phase II/III trials of golimumab in patients with moderate-to-severe UC. Visual predictive check was used for model evaluations. The ordered categorical approach is shown to be accurate and robust compared to the continuous approach. In addition, a disease progression model with an empirical bi-phasic rate of onset was found to be superior to the commonly used placebo model with one onset rate. An application of this modeling approach in guiding potential dose-adjustment was illustrated.