Skip Navigation and Go To Content

Houston-location seminar: On a novel deep neural network based mass imputation for data integration

When & Where

June 10, 2024
11:00 AM - 12:00 PM
Houston, RAS 102A or WebEx ( View in Google Map)

Contact

Event Description

Dr. Sixia Chen of the University of Oklahoma will be coming to our Houston Location on Monday, 6/10 to deliver a seminar titled “On a novel deep neural network based mass imputation for data integration.” He will be coming in person to RAS 102A at 11 am.

Webex link is in the event website field.

Password: aqDvuSVP358

Abstract

While probability samples are traditionally considered the gold standard for collecting high-quality data, the recent surge in the use of non-probability samples is notable. This shift is attributed to declining response rates, escalating survey costs, and the widespread availability of inexpensive and timely big non-probability data. However, relying solely on naive estimates derived from non-probability samples may expose analyses to substantial selection bias. Data integration techniques, which amalgamate information from both probability and non-probability samples, have proven highly effective in mitigating the selection bias inherent in non-probability samples. Established methods for data integration encompass mass imputation, propensity score weighting, calibration, and hybrid approaches, each contingent on specific model assumptions. In this paper, we address the challenge of model misspecification by introducing a novel deep neural network-based mass imputation method for data integration. Unlike existing deep neural network methods, our proposed approach is directly applicable for statistical inference, enhancing its utility in rigorous analyses. The paper provides asymptotic results to substantiate the method's validity. The results from both simulation studies and real-data applications underscore the robustness and effectiveness of our approach in improving the accuracy and reliability of estimates derived from integrated datasets.

Event Site Link

https://uthealth.webex.com/uthealth/j.php?MTID=mb2fd13973272da9c1a82cfb9e7906448

Additional Information

Houston-location seminar: On a novel deep neural network based mass imputation for data integration

{ "name":"Houston-location seminar: On a novel deep neural network based mass imputation for data integration", "description":"

Dr. Sixia Chen of the University of Oklahoma will be coming to our Houston Location on Monday, 6/10 to deliver a seminar titled “On a novel deep neural network based mass imputation for data integration.” He will be coming in person to RAS 102A at 11 am.

Webex link is in the event website field.

Password: aqDvuSVP358

Abstract

While probability samples are traditionally considered the gold standard for collecting high-quality data, the recent surge in the use of non-probability samples is notable. This shift is attributed to declining response rates, escalating survey costs, and the widespread availability of inexpensive and timely big non-probability data. However, relying solely on naive estimates derived from non-probability samples may expose analyses to substantial selection bias. Data integration techniques, which amalgamate information from both probability and non-probability samples, have proven highly effective in mitigating the selection bias inherent in non-probability samples. Established methods for data integration encompass mass imputation, propensity score weighting, calibration, and hybrid approaches, each contingent on specific model assumptions. In this paper, we address the challenge of model misspecification by introducing a novel deep neural network-based mass imputation method for data integration. Unlike existing deep neural network methods, our proposed approach is directly applicable for statistical inference, enhancing its utility in rigorous analyses. The paper provides asymptotic results to substantiate the method's validity. The results from both simulation studies and real-data applications underscore the robustness and effectiveness of our approach in improving the accuracy and reliability of estimates derived from integrated datasets.

\n\nhttps://uthealth.webex.com/uthealth/j.php?MTID=mb2fd13973272da9c1a82cfb9e7906448", "startDate":"2024-6-10", "endDate":"2024-6-10", "startTime":"11:00", "endTime":"12:00", "location":"Houston, RAS 102A or WebEx", "label":"Add to Calendar", "options":[ "Apple", "Google", "iCal", "Microsoft365", "MicrosoftTeams", "Outlook.com", "Yahoo" ], "timeZone":"America/Chicago", "trigger":"click", "inline":true, "listStyle":"modal", "iCalFileName":"Reminder-Event" }
Back