Inference and foundations

Skip to main content
Skip to footer

Language selection

Français

Search and menus

Search and menus

Search

Skip to filters. View results.

Results

All (92)

All (92) (0 to 10 of 92 results)

1. Dealing with undercoverage for non-probability survey samples
Articles and reports: 12-001-X202300200005
Description: Population undercoverage is one of the main hurdles faced by statistical analysis with non-probability survey samples. We discuss two typical scenarios of undercoverage, namely, stochastic undercoverage and deterministic undercoverage. We argue that existing estimation methods under the positivity assumption on the propensity scores (i.e., the participation probabilities) can be directly applied to handle the scenario of stochastic undercoverage. We explore strategies for mitigating biases in estimating the mean of the target population under deterministic undercoverage. In particular, we examine a split population approach based on a convex hull formulation, and construct estimators with reduced biases. A doubly robust estimator can be constructed if a followup subsample of the reference probability survey with measurements on the study variable becomes feasible. Performances of six competing estimators are investigated through a simulation study and issues which require further investigation are briefly discussed.
Release date: 2024-01-03
2. QR prediction for statistical data integration
Articles and reports: 12-001-X202300200009
Description: In this paper, we investigate how a big non-probability database can be used to improve estimates of finite population totals from a small probability sample through data integration techniques. In the situation where the study variable is observed in both data sources, Kim and Tam (2021) proposed two design-consistent estimators that can be justified through dual frame survey theory. First, we provide conditions ensuring that these estimators are more efficient than the Horvitz-Thompson estimator when the probability sample is selected using either Poisson sampling or simple random sampling without replacement. Then, we study the class of QR predictors, introduced by Särndal and Wright (1984), to handle the less common case where the non-probability database contains no study variable but auxiliary variables. We also require that the non-probability database is large and can be linked to the probability sample. We provide conditions ensuring that the QR predictor is asymptotically design-unbiased. We derive its asymptotic design variance and provide a consistent design-based variance estimator. We compare the design properties of different predictors, in the class of QR predictors, through a simulation study. This class includes a model-based predictor, a model-assisted estimator and a cosmetic estimator. In our simulation setups, the cosmetic estimator performed slightly better than the model-assisted estimator. These findings are confirmed by an application to La Poste data, which also illustrates that the properties of the cosmetic estimator are preserved irrespective of the observed non-probability sample.
Release date: 2024-01-03
3. The missing information principle ‒ A paradigm for analysis of messy sample survey data
Articles and reports: 12-001-X202300200018
Description: Sample surveys, as a tool for policy development and evaluation and for scientific, social and economic research, have been employed for over a century. In that time, they have primarily served as tools for collecting data for enumerative purposes. Estimation of these characteristics has been typically based on weighting and repeated sampling, or design-based, inference. However, sample data have also been used for modelling the unobservable processes that gave rise to the finite population data. This type of use has been termed analytic, and often involves integrating the sample data with data from secondary sources.
Alternative approaches to inference in these situations, drawing inspiration from mainstream statistical modelling, have been strongly promoted. The principal focus of these alternatives has been on allowing for informative sampling. Modern survey sampling, though, is more focussed on situations where the sample data are in fact part of a more complex set of data sources all carrying relevant information about the process of interest. When an efficient modelling method such as maximum likelihood is preferred, the issue becomes one of how it should be modified to account for both complex sampling designs and multiple data sources. Here application of the Missing Information Principle provides a clear way forward.
In this paper I review how this principle has been applied to resolve so-called “messy” data analysis issues in sampling. I also discuss a scenario that is a consequence of the rapid growth in auxiliary data sources for survey data analysis. This is where sampled records from one accessible source or register are linked to records from another less accessible source, with values of the response variable of interest drawn from this second source, and where a key output is small area estimates for the response variable for domains defined on the first source.
Release date: 2024-01-03
4. Bayes, buttressed by design-based ideas, is the best overarching paradigm for sample survey inference
Articles and reports: 12-001-X202200200001
Description:
Conceptual arguments and examples are presented suggesting that the Bayesian approach to survey inference can address the many and varied challenges of survey analysis. Bayesian models that incorporate features of the complex design can yield inferences that are relevant for the specific data set obtained, but also have good repeated-sampling properties. Examples focus on the role of auxiliary variables and sampling weights, and methods for handling nonresponse. The article offers ten top reasons for favoring the Bayesian approach to survey inference.

Release date: 2022-12-15
5. Statistical inference with non-probability survey samples
Articles and reports: 12-001-X202200200002
Description:
We provide a critical review and some extended discussions on theoretical and practical issues with analysis of non-probability survey samples. We attempt to present rigorous inferential frameworks and valid statistical procedures under commonly used assumptions, and address issues on the justification and verification of assumptions in practical applications. Some current methodological developments are showcased, and problems which require further investigation are mentioned. While the focus of the paper is on non-probability samples, the essential role of probability survey samples with rich and relevant information on auxiliary variables is highlighted.
Release date: 2022-12-15
6. Comments by Michael A. Bailey on “Statistical inference with non-probability survey samples” – Non-probability samples: An assessment and way forward
Articles and reports: 12-001-X202200200003
Description:
Non-probability surveys play an increasing role in survey research. Wu’s essay ably brings together the many tools available when assuming the non-response is conditionally independent of the study variable. In this commentary, I explore how to integrate Wu’s insights in a broader framework that encompasses the case in which non-response depends on the study variable, a case that is particularly dangerous in non-probabilistic polling.

Release date: 2022-12-15
7. Comments by Michael R. Elliott on “Statistical inference with non-probability survey samples”
Articles and reports: 12-001-X202200200004
Description:
This discussion attempts to add to Wu’s review of inference from non-probability samples, as well as to highlighting aspects that are likely avenues for useful additional work. It concludes with a call for an organized stable of high-quality probability surveys that will be focused on providing adjustment information for non-probability surveys.

Release date: 2022-12-15
8. Comments by Sharon L. Lohr on “Statistical inference with non-probability survey samples”
Articles and reports: 12-001-X202200200005
Description:
Strong assumptions are required to make inferences about a finite population from a nonprobability sample. Statistics from a nonprobability sample should be accompanied by evidence that the assumptions are met and that point estimates and confidence intervals are fit for use. I describe some diagnostics that can be used to assess the model assumptions, and discuss issues to consider when deciding whether to use data from a nonprobability sample.

Release date: 2022-12-15
9. Comments by Xiao-Li Meng on “Statistical inference with non-probability survey samples” – Miniaturizing data defect correlation: A versatile strategy for handling non-probability samples
Articles and reports: 12-001-X202200200006
Description:
Non-probability samples are deprived of the powerful design probability for randomization-based inference. This deprivation, however, encourages us to take advantage of a natural divine probability that comes with any finite population. A key metric from this perspective is the data defect correlation (ddc), which is the model-free finite-population correlation between the individual’s sample inclusion indicator and the individual’s attribute being sampled. A data generating mechanism is equivalent to a probability sampling, in terms of design effect, if and only if its corresponding ddc is of N-1/2 (stochastic) order, where N is the population size (Meng, 2018). Consequently, existing valid linear estimation methods for non-probability samples can be recast as various strategies to miniaturize the ddc down to the N-1/2 order. The quasi design-based methods accomplish this task by diminishing the variability among the N inclusion propensities via weighting. The super-population model-based approach achieves the same goal through reducing the variability of the N individual attributes by replacing them with their residuals from a regression model. The doubly robust estimators enjoy their celebrated property because a correlation is zero whenever one of the variables being correlated is constant, regardless of which one. Understanding the commonality of these methods through ddc also helps us see clearly the possibility of “double-plus robustness”: a valid estimation without relying on the full validity of either the regression model or the estimated inclusion propensity, neither of which is guaranteed because both rely on device probability. The insight generated by ddc also suggests counterbalancing sub-sampling, a strategy aimed at creating a miniature of the population out of a non-probability sample, and with favorable quality-quantity trade-off because mean-squared errors are much more sensitive to ddc than to the sample size, especially for large populations.

Release date: 2022-12-15
10. Comments by Zhonglei Wang and Jae Kwang Kim on “Statistical inference with non-probability survey samples”
Articles and reports: 12-001-X202200200007
Description:
Statistical inference with non-probability survey samples is a notoriously challenging problem in statistics. We introduce two new methods of nonparametric propensity score technique for weighting in the non-probability samples. One is the information projection approach and the other is the uniform calibration in the reproducing kernel Hilbert space.

Release date: 2022-12-15

Data (0)

Data (0) (0 results)

No content available at this time.

Analysis (92)

Analysis (92) (70 to 80 of 92 results)

71. The effect of record linkage errors on statistical inference in cohort mortality studies Archived
Articles and reports: 11-522-X20010016277
Description:
This paper discusses in detail issues dealing with the technical aspects of designing and conducting surveys. It is intended for an audience of survey methodologists.
The advent of computerized record-linkage methodology has facilitated the conduct of cohort mortality studies in which exposure data in one database are electronically linked with mortality data from another database. In this article, the impact of linkage errors on estimates of epidemiological indicators of risk, such as standardized mortality ratios and relative risk regression model parameters, is explored. It is shown that these indicators can be subject to bias and additional variability in the presence of linkage errors, with false links and non-links leading to positive and negative bias, respectively, in estimates of the standardized mortality ratio. Although linkage errors always increase the uncertainty in the estimates, bias can be effectively eliminated in the special case in which the false positive rate equals the false negative rate within homogeneous states defined by cross-classification of the covariates of interest.
Release date: 2002-09-12
72. Schooling, Literacy and Individual Earnings Archived
Articles and reports: 89-552-M2000007
Geography: Canada
Description:
This paper addresses the problem of statistical inference with ordinal variates and examines the robustness to alternative literacy measurement and scaling choices of rankings of average literacy and of estimates of the impact of literacy on individual earnings.
Release date: 2000-06-02
73. Meta analysis of bioassay data from the U.S. national toxicology program Archived
Articles and reports: 11-522-X19990015654
Description:
A meta analysis was performed to estimate the proportion of liver carcinogens, the proportion of chemicals carcinogenic at any site, and the corresponding proportion of anticarcinogens among chemicals tested in 397 long-term cancer bioassays conducted by the U.S. National Toxicology Program. Although the estimator used was negatively biased, the study provided persuasive evidence for a larger proportion of liver carcinogens (0.43,90%CI: 0.35,0.51) than was identified by the NTP (0.28). A larger proportion of chemicals carcinogenic at any site was also estimated (0.59,90%CI: 0.49,0.69) than was identified by the NTP (0.51), although this excess was not statistically significant. A larger proportion of anticarcinogens (0.66) was estimated than carcinogens (0.59). Despite the negative bias, it was estimated that 85% of the chemicals were either carcinogenic or anticarcinogenic at some site in some sex-species group. This suggests that most chemicals tested at high enough doses will cause some sort of perturbation in tumor rates.
Release date: 2000-03-02
74. Delineation of Canada's North: An Examination of the North-South Relationship in Canada Archived
Articles and reports: 92F0138M2000003
Description:
Statistics Canada's interest in a common delineation of the north for statistical analysis purposes evolved from research to devise a classification to further differentiate the largely rural and remote areas that make up 96% of Canada's land area. That research led to the establishment of the census metropolitan area and census agglomeration influenced zone (MIZ) concept. When applied to census subdivisions, the MIZ categories did not work as well in northern areas as in the south. Therefore, the Geography Division set out to determine a north-south divide that would differentiate the north from the south independent of any standard geographic area boundaries.
This working paper describes the methodology used to define a continuous line across Canada to separate the north from the south, as well as lines marking transition zones on both sides of the north-south line. It also describes the indicators selected to derive the north-south line and makes comparisons to alternative definitions of the north. The resulting classification of the north complements the MIZ classification. Together, census metropolitan areas, census agglomerations, MIZ and the North form a new Statistical Area Classification (SAC) for Canada.
Two related Geography working papers (catalogue no. 92F0138MPE) provide further details about the MIZ classification. Working paper no. 2000-1 (92F0138MPE00001) briefly describes MIZ and includes tables of selected socio-economic characteristics from the 1991 Census tabulated by the MIZ categories, and working paper no. 2000-2 (92F0138MPE00002) describes the methodology used to define the MIZ classification.
Release date: 2000-02-03
75. Comparative Study of Analytical Consumer Price Indexes for Different Subgroups of the Reference Population Archived
Articles and reports: 62F0014M1998013
Geography: Canada
Description:
The reference population for the Consumer Price Index (CPI) has been represented, since the 1992 updating of the basket of goods and services, by families and unattached individuals living in private urban or rural households. The official CPI is a measure of the average percentage change over time in the cost of a fixed basket of goods and services purchased by Canadian consumers.
Because of the broadly defined target population of the CPI, the measure has been criticised for failing to reflect the inflationary experiences of certain socio-economic groups. This study examines this question for three sub-groups of the reference population of the CPI. It is an extension of earlier studies on the subject done at Statistics Canada.
In this document, analytical consumer price indexes sub-group indexes are compared to the analytical index for the whole population calculated at the national geographic level.
The findings tend to point to those of earlier Statistics Canada studies on sub-groups in the CPI reference population. Those studies have consistently concluded that a consumer price index established for a given sub-group does not differ substantially from the index for the whole reference population.
Release date: 1999-05-13
76. Inverse sampling design algorithms Archived
Articles and reports: 12-001-X19970013101
Description:
In the main body of statistics, sampling is often disposed of by assuming a sampling process that selects random variables such that they are independent and identically distributed (IID). Important techniques, like regression and contingency table analysis, were developed largely in the IID world; hence, adjustments are needed to use them in complex survey settings. Rather than adjust the analysis, however, what is new in the present formulation is to draw a second sample from the original sample. In this second sample, the first set of selections are inverted, so as to yield at the end a simple random sample. Of course, to employ this two-step process to draw a single simple random sample from the usually much larger complex survey would be inefficient, so multiple simple random samples are drawn and a way to base inferences on them developed. Not all original samples can be inverted; but many practical special cases are discussed which cover a wide range of practices.
Release date: 1997-08-18
77. Variable selection for regression estimation in finite populations Archived
Articles and reports: 12-001-X19970013102
Description:
The selection of auxiliary variables is considered for regression estimation in finite populations under a simple random sampling design. This problem is a basic one for model-based and model-assisted survey sampling approaches and is of practical importance when the number of variables available is large. An approach is developed in which a mean squared error estimator is minimised. This approach is compared to alternative approaches using a fixed set of auxiliary variables, a conventional significance test criterion, a condition number reduction approach and a ridge regression approach. The proposed approach is found to perform well in terms of efficiency. It is noted that the variable selection approach affects the properties of standard variance estimators and thus leads to a problem of variance estimation.
Release date: 1997-08-18
78. A transformation method for finite population sampling calibrated with empirical likelihood Archived
Articles and reports: 12-001-X19960022980
Description:
In this paper, we study a confidence interval estimation method for a finite population average when some auxiliairy information is available. As demonstrated by Royall and Cumberland in a series of empirical studies, naive use of existing methods to construct confidence intervals for population averages may result in very poor conditional coverage probabilities, conditional on the sample mean of the covariate. When this happens, we propose to transform the data to improve the precision of the normal approximation. The transformed data are then used to make inference on the original population average, and the auxiliary information is incorporated into the inference directly, or by calibration with empirical likelihood. Our approach is design-based. We apply our approach to six real populations and find that when transformation is needed, our approach performs well compared to the usual regression method.
Release date: 1997-01-30
79. Fertility Projections for Canada, Provinces and Territories, 1993-2016 Archived
Articles and reports: 91F0015M1996001
Geography: Canada
Description:
This paper describes the methodology for fertility projections used in the 1993-based population projections by age and sex for Canada, provinces and territories, 1993-2016. A new version of the parametric model known as the Pearsonian Type III curve was applied for projecting fertility age pattern. The Pearsonian Type III model is considered as an improvement over the Type I used in the past projections. This is because the Type III curve better portrays both the distribution of the age-specific fertility rates and the estimates of births. Since the 1993-based population projections are the first official projections to incorporate the net census undercoverage in the population base, it has been necessary to recalculate fertility rates based on the adjusted population estimates. This recalculation resulted in lowering the historical series of age-specific and total fertility rates, 1971-1993. The three sets of fertility assumptions and projections were developed with these adjusted annual fertility rates.
It is hoped that this paper will provide valuable information about the technical and analytical aspects of the current fertility projection model. Discussions on the current and future levels and age pattern of fertility in Canada, provinces and territories are also presented in the paper.
Release date: 1996-08-02
80. Multiple sample estimation of population and census undercount in the presence of matching errors Archived
Articles and reports: 12-001-X199600114385
Description:
The multiple capture-recapture census is reconsidered by relaxing the traditional perfect matching assumption. We propose matching error models to characterize error-prone matching mechanisms. The observed data take the form of an incomplete 2^k contingency table with one missing cell and follow a multinomial distribution. We develop a procedure for the estimation of the population size. Our approach applies to both standard log-linear models for contingency tables and log-linear models for heterogeneity of catchability. We illustrate the method and estimation using a 1988 dress rehearsal study for the 1990 census conducted by the U.S. Bureau of the Census.
Release date: 1996-06-14

Reference (0)

Reference (0) (0 results)

No content available at this time.

Report a problem or mistake on this page

Date modified:: 2024-05-18