Search

In observational studies, one of the main difficulties consists in the comparison of treat- ment effects. In fact, receiving a treatment is not a “purely random” event, and there could be relevant differences between treatment groups. Propensity score is a popular tool to account for this source of bias. However, its use requires a careful modelization of the dependence relationships of the treatment on the covariates. In this work, we consider a general setting with multiple treatments and discrete multi-valued outcome. We propose to estimate the propensity score by using Bayesian Networks and, based on this, we develop an inferential methodology to evaluate the treatments effect. The performance of the pro- posed approach have been studied through a simulation study with very promising results.

This paper intends to contribute to an up-to-date discussion of dual frame designs in agricultural surveys. It starts by reviewing historical scenarios of applications to envision new perspectives, and ends by presenting a modern approach to the problem. A dual frame sampling design is proposed that has the appeal of relying on low-cost technological resources. The design has enough generality to allow for applications not only on agricultural but also on rural and environmental surveys, or any other survey related to the use of soil. Unbiased estimations based on domain and multiplicity approaches are presented and their major differences are discussed. Design parameters, design feasibility by different sample size allocations, as well as the statistical performance of several dual frame estimators are investigated using a Monte Carlo simulation study that is built on information from the Brazilian agricultural census of 2006 and FAO’s Global Strategy’s field experiences in the city of Goiana, Pernambuco. The results show dual frames present a gain in precision when compared to a single area frame survey. In addition, the choice of the best design and estimator depends upon scenarios with different types of allocation and different sizes of area frame segments.

A new class of sampling strategies is proposed that can be applied to population-based surveys targeting a rare trait that is unevenly spread over an area of interest. Our proposal is characterised by the ability to tailor the data collection to specific features and challenges of the survey at hand. It is based on integrating an adaptive component into a sequential selection, which aims both to intensify the detection of positive cases, upon exploiting the spatial clustering, and to provide a flexible framework to manage logistics and budget constraints. A class of estimators is also proposed to account for the selection bias, that are proved unbiased for the population mean (prevalence) as well as consistent and asymptotically Normal distributed. Unbiased variance estimation is also provided. A ready-to-implement weighting system is developed for estimation purposes. Two special strategies included in the proposed class are presented, that are based on the Poisson sampling and proved more efficient. The selection of primary sampling units is also illustrated for tuberculosis prevalence surveys, which are recommended in many countries and supported by the World Health Organisation as an emblematic example of the need for an improved sampling design. Simulation results are given in the tuberculosis application to illustrate the strengths and weaknesses of the proposed sequential adaptive sampling strategies with respect to traditional cross-sectional non-informative sampling as currently suggested by World Health Organisation guidelines.

The pursuance of gender equality has embraced a longstanding sta- tistical engendering process, to reflect women’s and men’s lives. In pursuing the 2030 Sustainability Development Goals (SDGs), the availability of high- quality gender-sensitive data has generated the current informative outburst. In the process, gender-sensitive data collection has departed from a mere dis- aggregation between men and women towards an unprecedented multifaceted informational spectrum. Methods for full exploiting gender-sensitive statis- tics, both standard and big data, though, faces some levels of criticality. The traditional descriptive linear combinations of a collection of simple indica- tors yields contradictory order results, whereas inference has so far privileged latent modelling only, holding several constraints. A novel statistical perspec- tive stems from recent developments of Multivariate Latent Markov Models (MLMMs), suitable to express a latent characteristic both in time and space. In addition to introducing covariates, on any measurement scales, not only in the structural model bul also the measurement one, MLMMs are innovative in so that they can handle a vast mass of data from very different sources. Thus they lead the way to an extensive investigation of the gender gap, account- ing for apparent contradictions in rankings and hence highlighting different paths, or “transition”, toward a more equitable society.

When observational studies are used to establish the causal effects of treatments, the estimated effect is affected by treatment selection bias. The inverse propensity score weight (IPSW) is often used to deal with such bias. However, IPSW requires strong assumptions whose misspecifications and strategies to correct the misspecifications were rarely studied. We present a bootstrap bias correction of IPSW (BC-IPSW) to improve the performance of propensity score in dealing with treatment selection bias in the presence of failure to the ignorability and overlap assumptions. The approach was motivated by a real observational study to explore the potential of anticoagulant treatment for reducing mortality in patients with end-stage renal disease. The benefit of the treatment to enhance survival was demonstrated; the suggested BC-IPSW method indicated a statistically significant reduction in mortality for patients receiving the treatment. Using extensive simulations, we show that BC-IPSW substantially reduced the bias due to the misspecification of the ignorability and overlap assumptions. Further, we showed that IPSW is still useful to account for the lack of treatment randomization, but its advantages are stringently linked to the satisfaction of ignorability, indicating that the existence of relevant though unmeasured or unused covariates can worsen the selection bias.

Sampling a rare and clustered trait in a finite population is challenging: traditional sampling designs usually require a large sample size in order to obtain reasonably accurate estimates, resulting in a considerable investment of resources in front of the detection of a small number of cases. A notable example is the case of WHO’s tuberculosis (TB) prevalence surveys, crucial for countries that bear a high TB burden, the prevalence of cases being still less than 1%. In the latest WHO guidelines, spatial patterns are not explicitly accounted for, with the risk of missing a large number of cases; moreover, cost and logistic constraints can pose further problems. After reviewing the methodology in use by WHO, the use of adaptive and sequential approaches is discussed as natural alternatives to improve over the limits of the current practice. A simulation study is presented to highlight possible advantages and limitations of these alternatives, and an integrated approach, combining both adaptive and sequential features in a single sampling strategy is advocated as a promising methodological perspective.

The effect of the practice of rounding non-integer weights in bootstrapping complex samples from finite populations is investigated by means of an extended simulation study. The extent to which rounding can interfere with basic Bootstrap principles as well as with the formal properties of the final Bootstrap estimates are evaluated. Indications and recommendations on application of this practice are discussed.

A class of generalized multiplicity-adjusted Horvitz–Thompson (GMHT) estimators was introduced by Singh and Mecatti (J. Official Stat. 27(4):633–650, 2011, JOS) to provide a unified and systematic approach to existing estimators via GMHT-Regression. The main purpose of this chapter is to present key observations that led to the development of the unified principled framework. These are based on the use of zero functions as predictors in regression which play a fundamental role in statistical estimation such as quasi-likelihood. The key observations are listed below. First, an optimal combination of two estimators of the same domain is equivalent to regression of the simple multiplicity-adjusted HT (i.e., average of the two estimators) on the zero function defined by the difference of two estimators. Second, there are two types of zero functions for each overlapping domain—one based on domain count estimates and the other based on domain total estimates. Some estimators use both types of zero functions as predictors in regression while others optimally combine first the two domain count estimates and then apply Hájek-type ratio adjustments for each sample to optimally estimated domain counts before regressing on domain total zero functions. Third, the regression need not be optimal as it can be based on a working covariance structure to obtain robust consistent estimates; this is in view of the observation that optimal regression estimators are typically unstable (i.e., with high coefficient of variation) for complex designs due to lack of adequate degrees of freedom for estimating regression parameters. Fourth, for any regression estimator, it may be better to use count zero functions via Hájek-adjustment first because the initial GMHT estimator may not be well correlated with count zero functions. Fifth, it might be preferable to use a suitable working covariance matrix than the optimal one in order to obtain a calibration form in addition to making the estimator

Writing papers is an essential part of the research process. Researchers have a professional obligation of disseminating their results, making them available for others to use to enhance common scientific knowledge. Besides the fun of sharing their own ideas and views, to publish is essential in order to actually have a scientific career. Although scientific writing certainly has its own conventions and standards, I suspect there is no a unique true recipe making the trick. As a matter of fact I do not have any. However my quite long time in the academic arena has given me a pretty clear idea about how I do and do not like things done. In this paper I will be giving my personal view and rules, in the hope that sharing my own experience would do some good to others as it did for me.

This book comprises a selection of the papers presented at the Third Italian Conference on Survey Methodology, ITACOSM 2013, held in Milan in June 2013 (http://www.statistica.unimib.it/itacosm13/). ITACOSM is the bi-annual international meeting of S2G, the Survey Sampling Group of the Italian Statistical Society. It is intended as a forum for scientific discussion on developments in the theory and application of survey sampling methodologies for the human and natural sciences. The volume gathers 14 contributions that address a wide range of sampling methods and techniques central to international statistical research. Three invited papers by keynote speakers at the conference open the book, offering state-of-the- art reviews of some hot topics in sampling theory: calibration for nonresponse treatment (P. S. Kott), quantile regression methods for small area estimation (M. Pratesi) and estimation from multiple frame surveys (A.C. Singh). Eleven research papers follow, carefully selected from both invited and contributed sessions of the conference and listed in alphabetical order by author. The contents of these papers range across mainstream topics in both sampling theory and applications. Theoretical contributions consider inference from complex samples, randomized response techniques, bootstrap methods, weighting, modeling, imputation and effective use of auxiliary information; applications cover a varied collection of subjects in official household surveys, Bayesian networks for measurement error, auditing, mixed-mode surveys, business and economic surveys, agricultural data and geostatistics. The book is intended to serve as a “hot topics reference” for survey sampling researchers, professionals and practitioners in a wide array of application fields. This book is published thanks to a grant by Marcello Fontanesi, past-Chancellor of the University of Milan-Bicocca.

The problem of bootstrapping the estimator's variance under a probability proportional to size design is examined. Focusing on the Horvitz-Thompson estimator, three pi PS-bootstrap algorithms are introduced with the purpose of both simplifying available procedures and of improving efficiency. Results from a simulation study using both natural and artificial data are presented in order to empirically investigate the properties of the provided bootstrap variance estimators.