Calibration and XGBoost reweighting to reduce coverage and non-response biases in overlapping panel surveys: application to the Healthcare and Social Survey
Abstract Background Surveys have been used worldwide to provide information on the COVID-19 pandemic impact so as to prepare and deliver an effective Public Health response. Overlapping panel surveys allow longitudinal estimates and more accurate cross-sectional estimates to be obtained thanks to th...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2024-02-01
|
| Series: | BMC Medical Research Methodology |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12874-024-02171-z |
| _version_ | 1797274195828146176 |
|---|---|
| author | Luis Castro María del Mar Rueda Carmen Sánchez-Cantalejo Ramón Ferri Andrés Cabrera-León |
| author_facet | Luis Castro María del Mar Rueda Carmen Sánchez-Cantalejo Ramón Ferri Andrés Cabrera-León |
| author_sort | Luis Castro |
| collection | DOAJ |
| description | Abstract Background Surveys have been used worldwide to provide information on the COVID-19 pandemic impact so as to prepare and deliver an effective Public Health response. Overlapping panel surveys allow longitudinal estimates and more accurate cross-sectional estimates to be obtained thanks to the larger sample size. However, the problem of non-response is particularly aggravated in the case of panel surveys due to population fatigue with repeated surveys. Objective To develop a new reweighting method for overlapping panel surveys affected by non-response. Methods We chose the Healthcare and Social Survey which has an overlapping panel survey design with measurements throughout 2020 and 2021, and random samplings stratified by province and degree of urbanization. Each measurement comprises two samples: a longitudinal sample taken from previous measurements and a new sample taken at each measurement. Results Our reweighting methodological approach is the result of a two-step process: the original sampling design weights are corrected by modelling non-response with respect to the longitudinal sample obtained in a previous measurement using machine learning techniques, followed by calibration using the auxiliary information available at the population level. It is applied to the estimation of totals, proportions, ratios, and differences between measurements, and to gender gaps in the variable of self-perceived general health. Conclusion The proposed method produces suitable estimators for both cross-sectional and longitudinal samples. For addressing future health crises such as COVID-19, it is therefore necessary to reduce potential coverage and non-response biases in surveys by means of utilizing reweighting techniques as proposed in this study. |
| first_indexed | 2024-03-07T14:55:51Z |
| format | Article |
| id | doaj.art-ce8fbc85af3141e68213ce0a203c2047 |
| institution | Directory Open Access Journal |
| issn | 1471-2288 |
| language | English |
| last_indexed | 2024-03-07T14:55:51Z |
| publishDate | 2024-02-01 |
| publisher | BMC |
| record_format | Article |
| series | BMC Medical Research Methodology |
| spelling | doaj.art-ce8fbc85af3141e68213ce0a203c20472024-03-05T19:28:34ZengBMCBMC Medical Research Methodology1471-22882024-02-0124111910.1186/s12874-024-02171-zCalibration and XGBoost reweighting to reduce coverage and non-response biases in overlapping panel surveys: application to the Healthcare and Social SurveyLuis Castro0María del Mar Rueda1Carmen Sánchez-Cantalejo2Ramón Ferri3Andrés Cabrera-León4Department of Statistics and Operational Research, University of GranadaDepartment of Statistics and Operational Research, University of GranadaDepartment of Public Health, Andalusian School of Public HealthDepartment of Statistics and Operational Research, University of GranadaDepartment of Public Health, Andalusian School of Public HealthAbstract Background Surveys have been used worldwide to provide information on the COVID-19 pandemic impact so as to prepare and deliver an effective Public Health response. Overlapping panel surveys allow longitudinal estimates and more accurate cross-sectional estimates to be obtained thanks to the larger sample size. However, the problem of non-response is particularly aggravated in the case of panel surveys due to population fatigue with repeated surveys. Objective To develop a new reweighting method for overlapping panel surveys affected by non-response. Methods We chose the Healthcare and Social Survey which has an overlapping panel survey design with measurements throughout 2020 and 2021, and random samplings stratified by province and degree of urbanization. Each measurement comprises two samples: a longitudinal sample taken from previous measurements and a new sample taken at each measurement. Results Our reweighting methodological approach is the result of a two-step process: the original sampling design weights are corrected by modelling non-response with respect to the longitudinal sample obtained in a previous measurement using machine learning techniques, followed by calibration using the auxiliary information available at the population level. It is applied to the estimation of totals, proportions, ratios, and differences between measurements, and to gender gaps in the variable of self-perceived general health. Conclusion The proposed method produces suitable estimators for both cross-sectional and longitudinal samples. For addressing future health crises such as COVID-19, it is therefore necessary to reduce potential coverage and non-response biases in surveys by means of utilizing reweighting techniques as proposed in this study.https://doi.org/10.1186/s12874-024-02171-zPublic healthCOVID-19Panel surveysSamplingMachine learningNon-response bias |
| spellingShingle | Luis Castro María del Mar Rueda Carmen Sánchez-Cantalejo Ramón Ferri Andrés Cabrera-León Calibration and XGBoost reweighting to reduce coverage and non-response biases in overlapping panel surveys: application to the Healthcare and Social Survey BMC Medical Research Methodology Public health COVID-19 Panel surveys Sampling Machine learning Non-response bias |
| title | Calibration and XGBoost reweighting to reduce coverage and non-response biases in overlapping panel surveys: application to the Healthcare and Social Survey |
| title_full | Calibration and XGBoost reweighting to reduce coverage and non-response biases in overlapping panel surveys: application to the Healthcare and Social Survey |
| title_fullStr | Calibration and XGBoost reweighting to reduce coverage and non-response biases in overlapping panel surveys: application to the Healthcare and Social Survey |
| title_full_unstemmed | Calibration and XGBoost reweighting to reduce coverage and non-response biases in overlapping panel surveys: application to the Healthcare and Social Survey |
| title_short | Calibration and XGBoost reweighting to reduce coverage and non-response biases in overlapping panel surveys: application to the Healthcare and Social Survey |
| title_sort | calibration and xgboost reweighting to reduce coverage and non response biases in overlapping panel surveys application to the healthcare and social survey |
| topic | Public health COVID-19 Panel surveys Sampling Machine learning Non-response bias |
| url | https://doi.org/10.1186/s12874-024-02171-z |
| work_keys_str_mv | AT luiscastro calibrationandxgboostreweightingtoreducecoverageandnonresponsebiasesinoverlappingpanelsurveysapplicationtothehealthcareandsocialsurvey AT mariadelmarrueda calibrationandxgboostreweightingtoreducecoverageandnonresponsebiasesinoverlappingpanelsurveysapplicationtothehealthcareandsocialsurvey AT carmensanchezcantalejo calibrationandxgboostreweightingtoreducecoverageandnonresponsebiasesinoverlappingpanelsurveysapplicationtothehealthcareandsocialsurvey AT ramonferri calibrationandxgboostreweightingtoreducecoverageandnonresponsebiasesinoverlappingpanelsurveysapplicationtothehealthcareandsocialsurvey AT andrescabreraleon calibrationandxgboostreweightingtoreducecoverageandnonresponsebiasesinoverlappingpanelsurveysapplicationtothehealthcareandsocialsurvey |