Interval-censored unimodal kernel density estimation via data sharpening

Becker, Devan; Braun, W. John; White, Bethany J. G.

doi:10.1080/00949655.2017.1308510

Cited by 4 publications

(3 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another method which applies a "data sharpening"-based approach to interval-censored data is also highly dependent on bandwidth selection, with suboptimal default performance in a nontrivial number of scenarios. 40 Overall, our proposed procedure contributes to the field of nonparametric survival analysis by achieving improved accuracy and predictive ability, especially when available survival data is limited by censoring, small sample size, low prevalence, or few repeated measurements per subject. We have demonstrated that smoothing NPMLE estimates by complexity-based penalized likelihood is a simple but powerful approach for improving survival & density estimation as well as exact time-to-event prediction (imputation).…”

Section: Discussionmentioning

confidence: 99%

“…One drawback in their proposed method for choosing a smoothing bandwidth, which uses the computationally expensive process of cross‐validation and can result in a poor choice of bandwidth. Another method which applies a “data sharpening”‐based approach to interval‐censored data is also highly dependent on bandwidth selection, with suboptimal default performance in a nontrivial number of scenarios 40 …”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Improved nonparametric penalized maximum likelihood estimation for arbitrarily censored survival data

Tubbs¹,

Thach²,

Sham

2022

Statistics in Medicine

View full text Add to dashboard Cite

Nonparametric maximum likelihood estimation encompasses a group of classic methods to estimate distribution-associated functions from potentially censored and truncated data, with extensive applications in survival analysis. These methods, including the Kaplan-Meier estimator and Turnbull's method, often result in overfitting, especially when the sample size is small. We propose an improvement to these methods by applying kernel smoothing to their raw estimates, based on a BIC-type loss function that balances the trade-off between optimizing model fit and controlling model complexity. In the context of a longitudinal study with repeated observations, we detail our proposed smoothing procedure and optimization algorithm. With extensive simulation studies over multiple realistic scenarios, we demonstrate that our smoothing-based procedure provides better overall accuracy in both survival function estimation and individual-level time-to-event prediction (imputation) by reducing overfitting. Our smoothing procedure decreases the bias (discrepancy between the estimated and true simulated survival function) using interval-censored data by up to 48% compared to the raw un-smoothed estimate, with similar improvements of up to 34% and 23% in within-sample and out-of-sample prediction, respectively. Our smoothing algorithm also demonstrates significant overall improvement across all three metrics when compared to a popular semiparametric B-splines estimation method. Finally, we apply our method to real data on censored breast cancer diagnosis, which similarly shows improvement when compared to empirical survival estimates from uncensored data. We provide an R package, SISE, for implementing our penalized likelihood method.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Improved nonparametric penalized maximum likelihood estimation for arbitrarily censored survival data

Tubbs¹,

Thach²,

Sham

2022

Statistics in Medicine

View full text Add to dashboard Cite

show abstract

“…Convergence of proposed algorithms is studied. In [4] data sharpening is proposed to increase robustness of a kernel density estimator to bandwidth misspecification and measurement errors. [19] used a maximum smoothed likelihood approach and a smoothing the (discrete) MLE of the distribution function approach.…”

Section: Introductionmentioning

confidence: 99%