Maximum likelihood versus multiple imputation for missing data in small longitudinal samples with nonnormality.

Shin, Tacksoo; Davison, Mark L.; Long, Jeffrey D.

doi:10.1037/met0000094

Cited by 107 publications

(65 citation statements)

References 138 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…An intent-to-treat perspective was used for all analyses. Maximum likelihood estimation using Mplus 8.0 (Muthén & Muthén, 2017) with auxiliary correlates (Graham, 2003) was used because research (Hayes & McArdle, 2017;Shin, Davison, & Long, 2017;Yuan-Wallentin, & Bentler, 2012) has shown multiple imputation does not properly handle missing data with smaller samples like that observed in our study, which was limited to two cases on the VABS II.…”

Section: Resultsmentioning

confidence: 99%

Randomized Control Trial of COMPASS for Improving Transition Outcomes of Students with Autism Spectrum Disorder

Ruble

McGrew

Toland

et al. 2018

J Autism Dev Disord

View full text Add to dashboard Cite

The postsecondary outcomes of individuals with autism spectrum disorder (ASD) are significantly worse than peers with other disabilities. One problem is the lack of empirically-supported transition planning interventions to guide services and help produce better outcomes. We applied an implementation science approach to adapt and modify an evidence-based consultation intervention originally tested with young children called the Collaborative Model for Promoting Competence and Success (COMPASS; Ruble et al., The collaborative model for promoting competence and success for students with ASD. Springer, New York, 2012a) and evaluate it for efficacy in a randomized controlled trial for transition-age youth. Results replicated findings with younger students with ASD that IEP outcomes were higher for COMPASS compared to the placebo control group (d = 2.1). Consultant fidelity was high and teacher adherence improved over time, replicating the importance of ongoing teacher coaching.

show abstract

Section: Resultsmentioning

confidence: 99%

Randomized Control Trial of COMPASS for Improving Transition Outcomes of Students with Autism Spectrum Disorder

Ruble

McGrew

Toland

et al. 2018

J Autism Dev Disord

View full text Add to dashboard Cite

show abstract

“…Given the random nature of the missing data, a path analysis using full information maximum likelihood estimation in AMOS version 23 (IBM Corp. Armonk, NY, USA) was chosen as a means to estimate cross-lagged and autoregressive pathways between variables in 6th class and 1st year [60]. Maximum likelihood estimation is recommended for longitudinal data with missing values that are MCAR as it uses all available data for each participant to estimate model pathways [61,62].…”

Section: Data Processingmentioning

confidence: 99%

What Keeps Them Physically Active? Predicting Physical Activity, Motor Competence, Health-Related Fitness, and Perceived Competence in Irish Adolescents after the Transition from Primary to Second-Level School

Britton

Issartel

Symonds

et al. 2020

IJERPH

View full text Add to dashboard Cite

Physical activity (PA) decreases with age. The school transition is noted for significant changes in PA behaviour. Motor competence (MC), health-related fitness (HRF), and perceived competence (PC) are generally positively associated with PA. The aim of this study was to examine longitudinal cross-lagged relationships between PA, MC, HRF, and PC across the school transition from final year of primary school to first year of second-level school in Irish youth. PA (accelerometery), object-control and locomotor MC (TGMD-III), PC (perceived athletic competence subscale of the Self-Perception Profile for Adolescents), and HRF (20 m shuttle run, horizontal jump, vertical jump, push-ups, curl-ups) were measured in final year of primary school (6th class) and first year of second-level school (1st year). In the sixth class, 261 participants (53% female; mean age 12.22 ± 0.48 years) were tested. In first year, 291 participants (48% female; mean age: 13.20 ± 0.39 years) were tested. In total, 220 participants were involved in the study at both timepoints. Cross-lagged regression in AMOS23, using full information maximum likelihood estimation, was conducted to test reciprocal and predictive pathways between variables. The full cross-lagged model showed acceptable fit (χ2 = 69.12, df = 8, p < 0.01, NFI = 0.93, CFI = 0.94). HRF was the strongest predictor of future PA (β = 0.353), and also predicted PC (β = 0.336) and MC (β = 0.163). Object-control MC predicted future PA (β = 0.192). Reciprocal relationships existed between object-control MC and PA, and between object-control MC and PC. HRF was the strongest predictor of PA. Object-control MC also predicted PA. PA promotion strategies should target the development of HRF and object-control MC in primary school to reduce the decline in PA frequently observed after the school transition.

show abstract

“…There is a rich body of simulation literature examining many different aspects of performance surrounding the LGM (see, e.g., Hertzog et al, 2008;Shin, Davison, & Long, 2017;Tong & Ke, 2016;Ye, 2016). One aspect that is commonly addressed is the performance of LGMs under small sample sizes (see, e.g., McNeish, 2016a(see, e.g., McNeish, , 2016b(see, e.g., McNeish, , 2017van de Schoot et al, 2015;Zondervan-Zwijnenburg et al, 2018).…”

Section: Latent Growth Models With a Distal Outcomementioning

confidence: 99%

Predicting a Distal Outcome Variable From a Latent Growth Model: ML versus Bayesian Estimation

Smid

Depaoli

Schoot

2019

Structural Equation Modeling: A Multidisciplinary Journal

View full text Add to dashboard Cite

Latent growth models (LGMs) with a distal outcome allow researchers to assess longer-term patterns, and to detect the need to start a (preventive) treatment or intervention in an early stage. The aim of the current simulation study is to examine the performance of an LGM with a continuous distal outcome under maximum likelihood (ML) and Bayesian estimation with default and informative priors, under varying sample sizes, effect sizes and slope variance values. We conclude that caution is needed when predicting a distal outcome from an LGM when the: (1) sample size is small; and (2) amount of variation around the latent slope is small, even with a large sample size. We recommend against the use of ML and Bayesian estimation with Mplus default priors in these situations to avoid severely biased estimates. Recommendations for substantive researchers working with LGMs with distal outcomes are provided based on the simulation results.

show abstract

Maximum likelihood versus multiple imputation for missing data in small longitudinal samples with nonnormality.

Cited by 107 publications

References 138 publications

Randomized Control Trial of COMPASS for Improving Transition Outcomes of Students with Autism Spectrum Disorder

Randomized Control Trial of COMPASS for Improving Transition Outcomes of Students with Autism Spectrum Disorder

What Keeps Them Physically Active? Predicting Physical Activity, Motor Competence, Health-Related Fitness, and Perceived Competence in Irish Adolescents after the Transition from Primary to Second-Level School

Predicting a Distal Outcome Variable From a Latent Growth Model: ML versus Bayesian Estimation

Contact Info

Product

Resources

About