Strategies for handling missing data in longitudinal studies with questionnaires

Nazanin Nooraee, Geert Molenberghs, Johan Ormel, Edwin R. van den Heuvel

Onderzoeksoutput: Bijdrage aan tijdschriftTijdschriftartikelAcademicpeer review

Uittreksel

Missing data methods, maximum likelihood estimation (MLE) and multiple imputation (MI), for longitudinal questionnaire data were investigated via simulation. Predictive mean matching (PMM) was applied at both item and scale levels, logistic regression at item level and multivariate normal imputation at scale level. We investigated a hybrid approach which is combination of MLE and MI, i.e. scales from the imputed data are eliminated if all underlying items were originally missing. Bias and mean square error (MSE) for parameter estimates were examined. ML seemed to provide occasionally the best results in terms of bias, but hardly ever on MSE. All imputation methods at the scale level and logistic regression at item level hardly ever showed the best performance. The hybrid approach is similar or better than its original MI. The PMM-hybrid approach at item level demonstrated the best MSE for most settings and in some cases also the smallest bias.

TaalEngels
Pagina's3415-3436
Aantal pagina's22
TijdschriftJournal of Statistical Computation and Simulation
Volume88
Nummer van het tijdschrift17
DOI's
StatusGepubliceerd - 22 nov 2018

Vingerafdruk

Data handling
Longitudinal Study
Missing Data
Mean square error
Questionnaire
Multiple Imputation
Hybrid Approach
Maximum likelihood estimation
Logistics
Imputation
Logistic Regression
Maximum Likelihood Estimation
Multivariate Normal
Strategy
Longitudinal study
Missing data
Estimate
Hybrid approach
Multiple imputation
Simulation

Trefwoorden

    Citeer dit

    @article{1a91420329d34c219c29a6d785ada231,
    title = "Strategies for handling missing data in longitudinal studies with questionnaires",
    abstract = "Missing data methods, maximum likelihood estimation (MLE) and multiple imputation (MI), for longitudinal questionnaire data were investigated via simulation. Predictive mean matching (PMM) was applied at both item and scale levels, logistic regression at item level and multivariate normal imputation at scale level. We investigated a hybrid approach which is combination of MLE and MI, i.e. scales from the imputed data are eliminated if all underlying items were originally missing. Bias and mean square error (MSE) for parameter estimates were examined. ML seemed to provide occasionally the best results in terms of bias, but hardly ever on MSE. All imputation methods at the scale level and logistic regression at item level hardly ever showed the best performance. The hybrid approach is similar or better than its original MI. The PMM-hybrid approach at item level demonstrated the best MSE for most settings and in some cases also the smallest bias.",
    keywords = "Fully conditional specification, latent variable models, maximum likelihood, multiple imputation",
    author = "Nazanin Nooraee and Geert Molenberghs and Johan Ormel and {van den Heuvel}, {Edwin R.}",
    year = "2018",
    month = "11",
    day = "22",
    doi = "10.1080/00949655.2018.1520854",
    language = "English",
    volume = "88",
    pages = "3415--3436",
    journal = "Journal of Statistical Computation and Simulation",
    issn = "0094-9655",
    publisher = "Taylor and Francis Ltd.",
    number = "17",

    }

    Strategies for handling missing data in longitudinal studies with questionnaires. / Nooraee, Nazanin; Molenberghs, Geert; Ormel, Johan; van den Heuvel, Edwin R.

    In: Journal of Statistical Computation and Simulation, Vol. 88, Nr. 17, 22.11.2018, blz. 3415-3436.

    Onderzoeksoutput: Bijdrage aan tijdschriftTijdschriftartikelAcademicpeer review

    TY - JOUR

    T1 - Strategies for handling missing data in longitudinal studies with questionnaires

    AU - Nooraee,Nazanin

    AU - Molenberghs,Geert

    AU - Ormel,Johan

    AU - van den Heuvel,Edwin R.

    PY - 2018/11/22

    Y1 - 2018/11/22

    N2 - Missing data methods, maximum likelihood estimation (MLE) and multiple imputation (MI), for longitudinal questionnaire data were investigated via simulation. Predictive mean matching (PMM) was applied at both item and scale levels, logistic regression at item level and multivariate normal imputation at scale level. We investigated a hybrid approach which is combination of MLE and MI, i.e. scales from the imputed data are eliminated if all underlying items were originally missing. Bias and mean square error (MSE) for parameter estimates were examined. ML seemed to provide occasionally the best results in terms of bias, but hardly ever on MSE. All imputation methods at the scale level and logistic regression at item level hardly ever showed the best performance. The hybrid approach is similar or better than its original MI. The PMM-hybrid approach at item level demonstrated the best MSE for most settings and in some cases also the smallest bias.

    AB - Missing data methods, maximum likelihood estimation (MLE) and multiple imputation (MI), for longitudinal questionnaire data were investigated via simulation. Predictive mean matching (PMM) was applied at both item and scale levels, logistic regression at item level and multivariate normal imputation at scale level. We investigated a hybrid approach which is combination of MLE and MI, i.e. scales from the imputed data are eliminated if all underlying items were originally missing. Bias and mean square error (MSE) for parameter estimates were examined. ML seemed to provide occasionally the best results in terms of bias, but hardly ever on MSE. All imputation methods at the scale level and logistic regression at item level hardly ever showed the best performance. The hybrid approach is similar or better than its original MI. The PMM-hybrid approach at item level demonstrated the best MSE for most settings and in some cases also the smallest bias.

    KW - Fully conditional specification

    KW - latent variable models

    KW - maximum likelihood

    KW - multiple imputation

    UR - http://www.scopus.com/inward/record.url?scp=85053540499&partnerID=8YFLogxK

    U2 - 10.1080/00949655.2018.1520854

    DO - 10.1080/00949655.2018.1520854

    M3 - Article

    VL - 88

    SP - 3415

    EP - 3436

    JO - Journal of Statistical Computation and Simulation

    T2 - Journal of Statistical Computation and Simulation

    JF - Journal of Statistical Computation and Simulation

    SN - 0094-9655

    IS - 17

    ER -