Surrogate Residuals for Discrete Choice Models.
Ontology highlight
ABSTRACT:
SUBMITTER: Cheng C
PROVIDER: S-EPMC8018589 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Ordinal outcomes are common in scientific research and everyday practice, and we often rely on regression models to make inference. A long-standing problem with such regression analyses is the lack of effective diagnostic tools for validating model assumptions. The difficulty arises from the fact that an ordinal variable has discrete values that are labeled with, but not, numerical values. The values merely represent ordered categories. In this paper, we propose a surrogate approach to defining residuals for an ordinal outcome Y. The idea is to define a continuous variable S as a "surrogate" of Y and then obtain residuals based on S. For the general class of cumulative link regression models, we study the residual's theoretical and graphical properties. We show that the residual has null properties similar to those of the common residuals for continuous outcomes. Our numerical studies demonstrate that the residual has power to detect misspecification with respect to 1) mean structures; 2) link functions; 3) heteroscedasticity; 4) proportionality; and 5) mixed populations. The proposed residual also enables us to develop numeric measures for goodness-of-fit using classical distance notions. Our results suggest that compared to a previously defined residual, our residual can reveal deeper insights into model diagnostics. We stress that this work focuses on residual analysis, rather than hypothesis testing. The latter has limited utility as it only provides a single p-value, whereas our residual can reveal what components of the model are misspecified and advise how to make improvements.
| S-EPMC6133273 | biostudies-literature
Project description:We describe a new residual for general regression models, defined as pr(Y* < y) - pr(Y* > y), where y is the observed outcome and Y* is a random variable from the fitted distribution. This probability-scale residual can be written as E {sign(y, Y*)} whereas the popular observed-minus-expected residual can be thought of as E(y - Y*). Therefore, the probability-scale residual is useful in settings where differences are not meaningful or where the expectation of the fitted distribution cannot be calculated. We present several desirable properties of the probability-scale residual that make it useful for diagnostics and measuring residual correlation, especially across different outcome types. We demonstrate its utility for continuous, ordered discrete, and censored outcomes, including current status data, and with various models including Cox regression, quantile regression, and ordinal cumulative probability models, for which fully specified distributions are not desirable or needed, and in some cases suitable residuals are not available. The residual is illustrated with simulated data and real datasets from HIV-infected patients on therapy in the southeastern United States and Latin America.
| S-EPMC5364820 | biostudies-literature
Project description:Nonmonotone missing data arise routinely in empirical studies of social and health sciences, and when ignored, can induce selection bias and loss of efficiency. In practice, it is common to account for nonresponse under a missing-at-random assumption which although convenient, is rarely appropriate when nonresponse is nonmonotone. Likelihood and Bayesian missing data methodologies often require specification of a parametric model for the full data law, thus a priori ruling out any prospect for semiparametric inference. In this paper, we propose an all-purpose approach which delivers semiparametric inferences when missing data are nonmonotone and not at random. The approach is based on a discrete choice model (DCM) as a means to generate a large class of nonmonotone nonresponse mechanisms that are nonignorable. Sufficient conditions for nonparametric identification are given, and a general framework for fully parametric and semiparametric inference under an arbitrary DCM is proposed. Special consideration is given to the case of logit discrete choice nonresponse model (LDCM) for which we describe generalizations of inverse-probability weighting, pattern-mixture estimation, doubly robust estimation and multiply robust estimation.
| S-EPMC8118571 | biostudies-literature
Project description:The proposed research contributes to our understanding of incorporating heterogeneity in discrete choice models with respect to exogenous variables and decision rules. Specifically, the proposed latent segmentation based mixed models segment population to different classes with their own decision rules while also incorporating unobserved heterogeneity within the segment level models. In our analysis, we choose to consider both random utility and random regret theories. Further, instead of assuming the number of segments (as 2), we conduct an exhaustive exploration with multiple segments across the two decision rules. The model estimation is conducted using a stated preference data from 695 commuter cyclists compiled through a web-based survey. The probabilistic allocation of respondents to different segments indicates that female commuter cyclists are more utility oriented; however, the majority of the commuter cyclist's choice pattern is consistent with regret minimization mechanism. Overall, cyclists' route choice decisions are influenced by roadway attributes, cycling infrastructure availability, pollution exposure, and travel time. The analysis approach also allows us to investigate time based trade-offs across cyclists belonging to different classes. Interestingly, we observe that the trade-off values in regret and utility based segments for roadway attributes are similar in magnitude; but the values differ greatly for cycling infrastructure and pollution exposure attributes, particularly for maximum exposure levels.
| S-EPMC6268012 | biostudies-literature
Project description:We use individual-level health facility choice data from urban Senegal to estimate consumer preferences for facility characteristics related to maternal health services. We find that consumers consider a large number of quality related facility characteristics, as well as travel costs, when making their health facility choice. In contrast to the typical assumption in the literature, our findings indicate that individuals frequently bypass the facility nearest their home. In light of this, we show that the mismeasured data used commonly in the literature produces biased preference estimates; most notably, the literature likely overestimates consumer distaste for travel.
| S-EPMC7062358 | biostudies-literature
Project description:We compare the performance of maximum likelihood (ML) and simulated method of moments (SMM) estimation for dynamic discrete choice models. We construct and estimate a simplified dynamic structural model of education that captures some basic features of educational choices in the United States in the 1980s and early 1990s. We use estimates from our model to simulate a synthetic dataset and assess the ability of ML and SMM to recover the model parameters on this sample. We investigate the performance of alternative tuning parameters for SMM.
| S-EPMC4610014 | biostudies-literature
Project description:Differentiated service delivery holds great promise for streamlining the delivery of health services for HIV. This study used a discrete choice experiment to assess preferences for differentiated HIV treatment delivery model characteristics among 500 virally suppressed adults on antiretroviral therapy in Harare, Zimbabwe. Treatment model characteristics included location, consultation type, healthcare worker cadre, operation times, visit frequency and duration, and cost. A mixed effects logit model was used for parameter estimates to identify potential preference heterogeneity among participants, and interaction effects were estimated for sex and age as potential sources of divergence in preferences. Results indicated that participants preferred health facility-based services, less frequent visits, individual consultations, shorter waiting times, lower cost and, delivered by respectful and understanding healthcare workers. Some preference heterogeneity was found, particularly for location of service delivery and group vs. individual models; however, this was not fully explained by sex and age characteristics of participants. In urban areas, facility-based models, such as the Fast Track model requiring less frequent clinic visits, are likely to better align with patient preferences than some of the other community-based or group models that have been implemented. As Zimbabwe scales up differentiated treatment models for stable patients, a clear understanding of patient preferences can help in designing services that will ensure optimal utilization and improve the efficiency of service delivery.
| S-EPMC7846512 | biostudies-literature
Project description:IntroductionTo improve retention on HIV treatment in Africa, public health programs are promoting a family of innovations to service delivery-referred to as "differentiated service delivery" (DSD) models-which seek to better meet the needs of both systems and patients by reducing unnecessary encounters, expanding access, and incorporating peers and patients in patient care. Data on the relative desirability of different models to target populations, which is currently sparse, can help guide prioritization of specific models during scale-up.MethodsWe conducted a discrete choice experiment to assess patient preferences for various characteristics of treatment services. Clinically stable people living with HIV were recruited from an HIV clinic in Kisumu, Kenya. We selected seven attributes of DSD models drawn from literature review and previous qualitative work. We created a balanced and orthogonal design to identify main term effects. A total of ten choice tasks were solicited per respondent. We calculated relative utility (RU) for each attribute level, a numerical representation of the strength of patient preference. Data were analyzed using a Hierarchical Bayesian model via Sawtooth Software.ResultsOne hundred and four respondents (37.5% men, 41.1 years mean age) preferred receiving care at a health facility, compared with home-delivery or a community meeting point (RU = 69.3, -16.2, and -53.1, respectively; p << 0.05); receiving those services from clinicians and pharmacists-as opposed to lay health workers or peers (RU = 21.5, 5.9, -24.5; p < 0.05); and preferred an individual support system over a group support system (RU = 15.0 and 4.2; p < 0.05). Likewise, patients strongly preferred longer intervals between both clinical reviews (RU = 40.1 and -50.7 for 6- and 1-month spacing, respectively; p < 0.05) and between ART collections (RU = 33.6 and -49.5 for 6- and1-month spacing, respectively; p < 0.05).ConclusionAlthough health systems find community- and peer-based DSD models attractive, clinically stable patients expressed a preference for facility-based care as long as clinical visits were extended to biannual. These data suggest that multi-month scripting and fast-track models best align with patient preferences, an insight which can help prioritize use of different DSD models in the region.
| S-EPMC8386850 | biostudies-literature
Project description:Introduction. The Dirichlet distribution has been proposed for representing preference heterogeneity, but there is limited evidence on its suitability for modeling population preferences on treatment benefits and risks. Methods. We conducted a simulation study to compare how the Dirichlet and standard discrete choice models (multinomial logit [MNL] and mixed logit [MXL]) differ in their convergence to stable estimates of population benefit-risk preferences. The source data consisted of individual-level tradeoffs from an existing 3-attribute patient preference study (N = 560). The Dirichlet population model was fit directly to the attribute weights in the source data. The MNL and MXL population models were fit to the outcomes of a simulated discrete choice experiment in the same sample of 560 patients. Convergence to the parameter values of the Dirichlet and MNL population models was assessed with sample sizes ranging from 20 to 500 (100 simulations per sample size). Model variability was also assessed with coefficient P values. Results. Population preference estimates of all models were very close to the sample mean, and the MNL and MXL models had good fit (McFadden's adjusted R2 = 0.12 and 0.13). The Dirichlet model converged reliably to within 0.05 distance of the population preference estimates with a sample size of 100, where the MNL model required a sample size of 240 for this. The MNL model produced consistently significant coefficient estimates with sample sizes of 100 and higher. Conclusion. The Dirichlet model is likely to have smaller sample size requirements than standard discrete choice models in modeling population preferences for treatment benefit-risk tradeoffs and is a useful addition to health preference analyst's toolbox.
| S-EPMC6843605 | biostudies-literature
Project description:Epigenomes are tissue specific and thus the choice of surrogate tissue can play a critical role in interpreting neonatal epigenome-wide association studies (EWAS) and in their extrapolation to target tissue. To develop a better understanding of the link between tissue specificity and neonatal EWAS, and the contributions of genotype and prenatal factors, we compared genome-wide DNA methylation of cord tissue and cord blood, two of the most accessible surrogate tissues at birth.In 295 neonates, DNA methylation was profiled using Infinium HumanMethylation450 beadchip arrays. Sites of inter-individual variability in DNA methylation were mapped and compared across the two surrogate tissues at birth, i.e., cord tissue and cord blood. To ascertain the similarity to target tissues, DNA methylation profiles of surrogate tissues were compared to 25 primary tissues/cell types mapped under the Epigenome Roadmap project. Tissue-specific influences of genotype on the variable CpGs were also analyzed. Finally, to interrogate the impact of the in utero environment, EWAS on 45 prenatal factors were performed and compared across the surrogate tissues.Neonatal EWAS results were tissue specific. In comparison to cord blood, cord tissue showed higher inter-individual variability in the epigenome, with a lower proportion of CpGs influenced by genotype. Both neonatal tissues were good surrogates for target tissues of mesodermal origin. They also showed distinct phenotypic associations, with effect sizes of the overlapping CpGs being in the same order of magnitude.The inter-relationship between genetics, prenatal factors and epigenetics is tissue specific, and requires careful consideration in designing and interpreting future neonatal EWAS.This birth cohort is a prospective observational study, designed to study the developmental origins of health and disease, and was retrospectively registered on 1 July 2010 under the identifier NCT01174875 .
| S-EPMC5715509 | biostudies-literature