東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

An Inspection of Measurement Error I...

Miller, Joseph E.,

Linked to FindBook

Google Book

Amazon

博客來

An Inspection of Measurement Error Inherited in Time-varying Latent Covariate Proxies within Marginal Structural Models /

Record Type:	Electronic resources : Monograph/item
Title/Author:	An Inspection of Measurement Error Inherited in Time-varying Latent Covariate Proxies within Marginal Structural Models // Joseph E Miller.
Author:	Miller, Joseph E.,
Description:	1 electronic resource (98 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 86-04, Section: A.
Contained By:	Dissertations Abstracts International86-04A.
Subject:	Statistics. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31690617
ISBN:	9798384442509

An Inspection of Measurement Error Inherited in Time-varying Latent Covariate Proxies within Marginal Structural Models /
Miller, Joseph E.,

An Inspection of Measurement Error Inherited in Time-varying Latent Covariate Proxies within Marginal Structural Models /Joseph E Miller. - 1 electronic resource (98 pages)

Source: Dissertations Abstracts International, Volume: 86-04, Section: A.

In longitudinal studies, the estimation of the average treatment effect from time varying treatment assumes that all covariates are measured without error within the potential outcomes framework. However, this assumption is rarely met in social science literature. Time varying latent covariates represent unobserved variables which can be incorporated in Marginal Structural Models and Propensity Score estimation through proxies (e.g., indicators and factors scores) to account for confounding on the causal effect of treatment. Traditionally, applied statisticians could use a factor score generated under statistical assumptions (e.g., composite sum score, regression factor score) as latent covariate proxies when estimating treatment effect. Where some research has consistently shown differences between the different factor score generating methods (Rakyov, 2012); a general consensus has found that there is little to no difference between these methods (Whittaker, 2020). Recently, there has been development of a factor score estimated under causal inference assumptions (i.e., the exposure confounder joint distribution) called the inclusive factor score (Nguyen & Stuart, 2020). This new causally generated factor score has been shown to outperform and mitigate the inherit measurement error found in latent covariate proxies. However, this leaves a gap in the literature as the research is confined to time-fixed treatment alone with no recommendations on best practices for the inclusion of multiple exposure variables. Thus, a systematic review and examination of latent covariate proxies, time-varying treatment models and their respective time-varying covariates with measurement error is needed. This study examines such mismeasurement systematically through Monte Carlo simulations when correction methods are and are not used. Here, the outcome of interest focused on the potential bias in the estimation of the average treatment effect for different treatment histories. Variation in simulation conditions reflect parameter issues in measurement error as it relates to reliability, treatment group size, type of latent covariate proxy, degree of effect, and other contributing factors to propensity score models and marginal structural models. Overall, the latent covariate proxy approaches generally resulted in increased bias and less prediction accuracy as measurement error in the latent covariate increased. Unequal treatment and control group sizes was a large contributing factor among manipulated factors and resulted in increased bias and prediction accuracy as compared to conditions with equal control and treatment group sizes. Overall, the inclusive factor score that accounts for treatment at all time points performed more optimally than the comparison latent covariate proxy methods. Limitations and implications for researchers are discussed.

English

ISBN: 9798384442509Subjects--Topical Terms:

517247
Statistics.
Subjects--Index Terms:

Latent covariate proxy

An Inspection of Measurement Error Inherited in Time-varying Latent Covariate Proxies within Marginal Structural Models /
LDR:04513nmm a22004573i 4500 001 2430827
005 20250603111625.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798384442509
035 $a (MiAaPQD)AAI31690617
035 $a (MiAaPQD)123vireo24464Miller
035 $a AAI31690617
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Miller, Joseph E., $e author. $3 3819341
245 1 3 $a An Inspection of Measurement Error Inherited in Time-varying Latent Covariate Proxies within Marginal Structural Models / $c Joseph E Miller.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (98 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-04, Section: A.
500 $a Advisors: Whittaker, Tiffany A.; Liu, Xiao Committee members: Keller, Brian; Lewis, Nancy Scammacca.
502 $b Ph.D. $c The University of Texas at Austin $d 2024.
520 $a In longitudinal studies, the estimation of the average treatment effect from time varying treatment assumes that all covariates are measured without error within the potential outcomes framework. However, this assumption is rarely met in social science literature. Time varying latent covariates represent unobserved variables which can be incorporated in Marginal Structural Models and Propensity Score estimation through proxies (e.g., indicators and factors scores) to account for confounding on the causal effect of treatment. Traditionally, applied statisticians could use a factor score generated under statistical assumptions (e.g., composite sum score, regression factor score) as latent covariate proxies when estimating treatment effect. Where some research has consistently shown differences between the different factor score generating methods (Rakyov, 2012); a general consensus has found that there is little to no difference between these methods (Whittaker, 2020). Recently, there has been development of a factor score estimated under causal inference assumptions (i.e., the exposure confounder joint distribution) called the inclusive factor score (Nguyen & Stuart, 2020). This new causally generated factor score has been shown to outperform and mitigate the inherit measurement error found in latent covariate proxies. However, this leaves a gap in the literature as the research is confined to time-fixed treatment alone with no recommendations on best practices for the inclusion of multiple exposure variables. Thus, a systematic review and examination of latent covariate proxies, time-varying treatment models and their respective time-varying covariates with measurement error is needed. This study examines such mismeasurement systematically through Monte Carlo simulations when correction methods are and are not used. Here, the outcome of interest focused on the potential bias in the estimation of the average treatment effect for different treatment histories. Variation in simulation conditions reflect parameter issues in measurement error as it relates to reliability, treatment group size, type of latent covariate proxy, degree of effect, and other contributing factors to propensity score models and marginal structural models. Overall, the latent covariate proxy approaches generally resulted in increased bias and less prediction accuracy as measurement error in the latent covariate increased. Unequal treatment and control group sizes was a large contributing factor among manipulated factors and resulted in increased bias and prediction accuracy as compared to conditions with equal control and treatment group sizes. Overall, the inclusive factor score that accounts for treatment at all time points performed more optimally than the comparison latent covariate proxy methods. Limitations and implications for researchers are discussed.
546 $a English
590 $a School code: 0227
650 4 $a Statistics. $3 517247
650 4 $a Educational tests & measurements. $3 3168483
650 4 $a Applied mathematics. $3 2122814
653 $a Latent covariate proxy
653 $a Monte Carlo simulations
653 $a Marginal structural models
653 $a Causal inference
653 $a Prediction accuracy
690 $a 0463
690 $a 0288
690 $a 0364
710 2 $a The University of Texas at Austin. $b Educational Psychology. $e degree granting institution. $3 3819342
720 1 $a Whittaker, Tiffany A. $e degree supervisor.
720 1 $a Liu, Xiao $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-04A.
790 $a 0227
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31690617