Bootstrapping correlation coefficients using univariate and bivariate sampling.

Lee, Won‐Chan; Rodgers, Joseph Lee

doi:10.1037/1082-989x.3.1.91

Cited by 49 publications

(54 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Generally, the surrogate data method with percentile intervals outperformed the other tests in terms of better control of Type I error and higher statistical power. This finding is generally in accordance with the study on tests of correlations for complete but independent data (Lee & Rodgers, 1998). Additionally, the use of BCa intervals resulted in less satisfactory control of Type I error rates and substantial decrease in power, suggesting not using BCa intervals together with the surrogate data method.…”

Section: Discussionsupporting

confidence: 79%

“…A related study on tests of correlations for independent data showed that the univariate bootstrap which resamples under the null outperformed the bivariate bootstrap which resamples under the alternative when testing correlations (Lee & Rodgers, 1998). In addition, among the several interval construction methods for resampling techniques, bias-corrected and accelerated (BCa) intervals are often recommended because they are second-order accurate, as opposed to the first-order accurate percentile intervals (Efron & Tibshirani, 1994;p.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Testing autocorrelation and partial autocorrelation: Asymptotic methods versus resampling techniques

Zhang

2017

Brit J Math & Statis

View full text Add to dashboard Cite

Autocorrelation and partial autocorrelation, which provide a mathematical tool to understand repeating patterns in time series data, are often used to facilitate the identification of model orders of time series models (e.g., moving average and autoregressive models). Asymptotic methods for testing autocorrelation and partial autocorrelation such as the 1/T approximation method and the Bartlett's formula method may fail in finite samples and are vulnerable to non-normality. Resampling techniques such as the moving block bootstrap and the surrogate data method are competitive alternatives. In this study, we use a Monte Carlo simulation study and a real data example to compare asymptotic methods with the aforementioned resampling techniques. For each resampling technique, we consider both the percentile method and the bias-corrected and accelerated method for interval construction. Simulation results show that the surrogate data method with percentile intervals yields better performance than the other methods. An R package pautocorr is used to carry out tests evaluated in this study.

show abstract

Section: Discussionsupporting

confidence: 79%

Section: Introductionmentioning

confidence: 99%

Testing autocorrelation and partial autocorrelation: Asymptotic methods versus resampling techniques

Zhang

2017

Brit J Math & Statis

View full text Add to dashboard Cite

show abstract

“…Calculating an average similarity score that included each trainee's similarity to every other trainee is statistically problematic, because it mixes within and between subject data and thus violates the assumption of independence of observations. Therefore, we wrote a program in MatLab# (1996) that applied a variation of bootstrapping similar to the approximate randomization technique, to estimate an empirical sampling distribution from our data (Lee and Rodgers, 1998;Rasmussen, 1989). Speci®cally, for each group, we (1) randomly selected pairs of participants, without replacement, until we had exhausted the pool; (2) tallied their pair-wise similarity scores; and (3) calculated an average similarity score.…”

Section: Similarity Among Participantsmentioning

confidence: 99%

Measuring teamwork mental models to support training needs assessment, development, and evaluation: two empirical studies†

et al. 2001

View full text Add to dashboard Cite

SummaryThe present paper reports data from two studies that utilized a card sorting approach to measuring mental model similarity in naturalistic training environments. Results from the ®rst study indicated that higher ranking navy personnel held mental models of teamwork that were more similar to an empirically derived model of expert team performance than lower ranking personnel. Furthermore, comparisons of mental model similarity within groups of high and low ranking trainees and within groups of high and low experience trainees indicated greater similarity between those of higher rank and between those with greater experience. The second study tested the effects of a computer-based training (CBT) strategy that was designed to develop teamwork mental models that were more similar to the`expert model' described in Study 1. Using the same card sorting approach, positive training effects were demonstrated on similarity to the expert model, similarity to other trainees, and consistency.

show abstract

“…Previous research suggested that none of these approaches provides a perfect solution to the problem of nonnormality with correlated data. These methods can slightly inflate Type I error (Beasley et al, 2007;Bishara & Hittner, 2012;Lee & Rodgers, 1998;Rasmussen, 1987;Strube, 1988). Additionally, their 95 % confidence intervals can have actual coverage rates that range from 91 % to 99 % (Puth et al, 2014(Puth et al, , 2015.…”

Section: Bootstrap Methodsmentioning

confidence: 99%

Confidence intervals for correlations when data are not normal

2016

View full text Add to dashboard Cite

With nonnormal data, the typical confidence interval of the correlation (Fisher z') may be inaccurate. The literature has been unclear as to which of several alternative methods should be used instead, and how extreme a violation of normality is needed to justify an alternative. Through Monte Carlo simulation, 11 confidence interval methods were compared, including Fisher z', two Spearman rank-order methods, the Box-Cox transformation, rank-based inverse normal (RIN) transformation, and various bootstrap methods. Nonnormality often distorted the Fisher z' confidence interval-for example, leading to a 95 % confidence interval that had actual coverage as low as 68 %. Increasing the sample size sometimes worsened this problem. Inaccurate Fisher z' intervals could be predicted by a sample kurtosis of at least 2, an absolute sample skewness of at least 1, or significant violations of normality hypothesis tests. Only the Spearman rankorder and RIN transformation methods were universally robust to nonnormality. Among the bootstrap methods, an observed imposed bootstrap came closest to accurate coverage, though it often resulted in an overly long interval. The results suggest that sample nonnormality can justify avoidance of the Fisher z' interval in favor of a more robust alternative. R code for the relevant methods is provided in supplementary materials.

show abstract

Bootstrapping correlation coefficients using univariate and bivariate sampling.

Cited by 49 publications

References 22 publications

Testing autocorrelation and partial autocorrelation: Asymptotic methods versus resampling techniques

Testing autocorrelation and partial autocorrelation: Asymptotic methods versus resampling techniques

Measuring teamwork mental models to support training needs assessment, development, and evaluation: two empirical studies†

Confidence intervals for correlations when data are not normal

Contact Info

Product

Resources

About