Statistical Theory Commons^™

Caution For Software Use Of New Statistical Methods (R), Akiva J. Lorenz, Barry S. Markman, Shlomo Sawilowsky

Journal of Modern Applied Statistical Methods

Open source programming languages such as R allow statisticians to develop and rapidly disseminate advanced procedures, but sometimes at the expense of a proper vetting process. A new example is the least trimmed squares regression available in R’s lqs() in the MASS library. It produces pretty regression lines, particularly in the presence of outliers. However, this procedure lacks a defined standard error, and thus it should be avoided.

Inferences About The Skipped Correlation Coefficient: Dealing With Heteroscedasticity And Non-Normality, Rand Wilcox

Journal of Modern Applied Statistical Methods

A common goal is testing the hypothesis that Pearson’s correlation is zero and typically this is done based on Student’s T test. There are, however, several well-known concerns. First, Student’s T is sensitive to heteroscedasticity. That is, when it rejects, it is reasonable to conclude that there is dependence, but in terms of making a decision about the strength of the association, it is unsatisfactory. Second, Pearson’s correlation is not robust: it can poorly reflect the strength of the association. Even a single outlier can have a tremendous impact on the usual estimate of Pearson’s correlation, which can result in …

Front Matter, Jmasm Editors

Journal of Modern Applied Statistical Methods

.

Jmasm34: Two Group Program For Cohen's D, Hedges’ G, Η2, Radj2, Ω2, Ɛ2, Confidence Intervals, And Power, David A. Walker

Journal of Modern Applied Statistical Methods

The purpose of this research is to provide an application for users interested in a SPSS syntax program to determine an array of commonly-employed effect sizes and confidence intervals not readily available in SPSS functionality, such as the standardized mean difference and r-related squared indices, for a between-group design.

Monte Carlo Comparison Of The Parameter Estimation Methods For The Two-Parameter Gumbel Distribution, Demet Aydin, Birdal Şenoğlu

Journal of Modern Applied Statistical Methods

The performances of the seven different parameter estimation methods for the Gumbel distribution are compared with numerical simulations. Estimation methods used in this study are the method of moments (ME), the method of maximum likelihood (ML), the method of modified maximum likelihood (MML), the method of least squares (LS), the method of weighted least squares (WLS), the method of percentile (PE) and the method of probability weighted moments (PWM). Performance of the estimators is compared with respect to their biases, MSE and deficiency (Def) values via Monte-Carlo simulation. A Monte Carlo Simulation study showed that the method of PWM was …

Contrails: Causal Inference Using Propensity Scores, Dean S. Barron

Journal of Modern Applied Statistical Methods

Contrails are clouds caused by airplane exhausts, which geologists contend decrease daily temperature ranges on Earth. Following the 2001 World Trade Center attack, cancelled domestic flights triggered the first absence of contrails in decades. Resultant exceptional data capacitated causal inference analysis by propensity score matching. Estimated contrail effect was 6.8981°F.

Two Stage Robust Ridge Method In A Linear Regression Model, Adewale Folaranmi Lukman, Oyedeji Isola Osowole, Kayode Ayinde

Journal of Modern Applied Statistical Methods

Two Stage Robust Ridge Estimators based on robust estimators M, MM, S, LTS are examined in the presence of autocorrelation, multicollinearity and outliers as alternative to Ordinary Least Square Estimator (OLS). The estimator based on S estimator performs better. Mean square error was used as a criterion for examining the performances of these estimators.

The Bayes Factor For Case-Control Studies With Misclassified Data, Tzesan Lee

Journal of Modern Applied Statistical Methods

The question of how to test if collected data for a case-control study are misclassified was investigated. A mixed approach was employed to calculate the Bayes factor to assess the validity of the null hypothesis of no-misclassification. A real-world data set on the association between lung cancer and smoking status was used as an example to illustrate the proposed method.

Resolving The Issue Of How Reliability Is Related To Statistical Power: Adhering To Mathematical Definitions, Donald W. Zimmerman, Bruno D. Zumbo

Journal of Modern Applied Statistical Methods

Reliability in classical test theory is a population-dependent concept, defined as a ratio of true-score variance and observed-score variance, where observed-score variance is a sum of true and error components. On the other hand, the power of a statistical significance test is a function of the total variance, irrespective of its decomposition into true and error components. For that reason, the reliability of a dependent variable is a function of the ratio of true-score variance and observed-score variance, whereas statistical power is a function of the sum of the same two variances. Controversies about how reliability is related to statistical …

In (Partial) Defense Of .05, Thomas R. Knapp

Journal of Modern Applied Statistical Methods

Researchers are frequently chided for choosing the .05 alpha level as the determiner of statistical significance (or non-significance). A partial justification is provided.

Vol. 14, No. 2 (Full Issue), Jmasm Editors

Journal of Modern Applied Statistical Methods

.

Semi-Parametric Non-Proportional Hazard Model With Time Varying Covariate, Kazeem A. Adeleke, Alfred A. Abiodun, R. A. Ipinyomi

Journal of Modern Applied Statistical Methods

The application of survival analysis has extended the importance of statistical methods for time to event data that incorporate time dependent covariates. The Cox proportional hazards model is one such method that is widely used. An extension of the Cox model with time-dependent covariates was adopted when proportionality assumption are violated. The purpose of this study is to validate the model assumption when hazard rate varies with time. This approach is applied to model data on duration of infertility subject to time varying covariate. Validity is assessed by a set of simulation experiments and results indicate that a non proportional …

Bayesian Analysis Under Progressively Censored Rayleigh Data, Gyan Prakash

Journal of Modern Applied Statistical Methods

The one-parameter Rayleigh model is considered as an underlying model for evaluating the properties of Bayes estimator under Progressive Type-II right censored data. The One‑Sample Bayes prediction bound length (OSBPBL) is also measured. Based on two different asymmetric loss functions a comparative study presented for Bayes estimation. A simulation study was used to evaluate their comparative properties.

Statistical Modeling Of Migration Attractiveness Of The Eu Member States, Tatiana Tikhomirova, Yulia Lebedeva

Journal of Modern Applied Statistical Methods

Identifying the relationship between the migration attractiveness of the European Union countries and their level of socio-economic development is investigated. An approach is proposed identify influences on migration socio-economic characteristics, by aggregating and reducing their diversity, and substantiating the cause-and-effect relationships of the studied phenomenon. A stable classification of countries scheme is developed according to the attractiveness of migration on aggregate factors, and then an econometric model of a binary choice using panel data for 2008-2010 was applying, quantifying the impact of aggregate designed factors on immigration and emigration.

The Distribution Of The Inverse Square Root Transformed Error Component Of The Multiplicative Time Series Model, Bright F. Ajibade, Chinwe R. Nwosu, J. I. Mbegdu

Journal of Modern Applied Statistical Methods

The probability density function, mean and variance of the inverse square-root transformed left-truncated N(1,σ²) error component e^*_t(=1/ √e_t) of the multiplicative time series model were established. A comparison of key-statistical properties of e^*_t and e_t confirmed normality with mean 1 but with Var(e^*_t) ≈1/4Var(e_t) when σ≤0.14. Hence σ≤0.14 is the required condition for successful transformation.

Approaches For Detection Of Unstable Processes: A Comparative Study, Yerriswamy Wooluru, D. R. Swamy, P. Nagesh

Journal of Modern Applied Statistical Methods

A process is stable only when parameters of the distribution of a process or product characteristic remain same over time. Only a stable process has the ability to perform in a predictable manner over time. Statistical analysis of process data usually assume that data are obtained from stable process. In the absence of control charts, the hypothesis of process stability is usually assessed by visual examination of the pattern in the run chart. In this paper appropriate statistical approaches have been adopted to detect instability in the process and compared their performance with the run chart of considerably shorter length …

A Robust Panel Unit Root Test In The Presence Of Cross Sectional Dependence, Nurul Sima Mohamad Shariff, Nor Aishah Hamzah

Journal of Modern Applied Statistical Methods

Problems arise in testing the stationarity of the panel in the presence of cross sectional dependence and outliers. The currently available panel unit root tests are very much affected by the presence of outliers. As such, this article introduces an alternative test which is robust to outliers and cross sectional dependence. The performance and robustness of the proposed test is discussed and comparisons are made to the existing tests via simulation studies.

C-Learning: A New Classification Framework To Estimate Optimal Dynamic Treatment Regimes, Baqun Zhang, Min Zhang

The University of Michigan Department of Biostatistics Working Paper Series

Personalizing treatment to accommodate patient heterogeneity and the evolving nature of a disease over time has received considerable attention lately. A dynamic treatment regime is a set of decision rules, each corresponding to a decision point, that determine that next treatment based on each individual’s own available characteristics and treatment history up to that point. We show that identifying the optimal dynamic treatment regime can be recast as a sequential classification problem and is equivalent to sequentially minimizing a weighted expected misclassification error. This general classification perspective targets the exact goal of optimally individualizing treatments and is new and fundamentally …

A Study Of The Parametric And Nonparametric Linear-Circular Correlation Coefficient, Robin Tu

Statistics

Circular statistics are specialized statistical methods that deal specifically with directional data. Data that is angular require specialized techniques due to the modulo 2π (in radians) or modulo 360^◦ (in degrees) nature of angles.

Correlation, typically in terms of Pearson’s correlation coefficient, is a measure of association between two linear random variables x and y. In this paper, the specific circular technique of the parametric and nonparametric linear-circular correlation coefficient will be explored where correlation is no longer between two linear variables x and y, but between a linear random variable x and circular random variable θ.

A simulation …

Applying Penalized Binary Logistic Regression With Correlation Based Elastic Net For Variables Selection, Zakariya Yahya Algamal, Muhammad Hisyam Lee

Journal of Modern Applied Statistical Methods

Reduction of the high dimensional classification using penalized logistic regression is one of the challenges in applying binary logistic regression. The applied penalized method, correlation based elastic penalty (CBEP), was used to overcome the limitation of LASSO and elastic net in variable selection when there are perfect correlation among explanatory variables. The performance of the CBEP was demonstrated through its application in analyzing two well-known high dimensional binary classification data sets. The CBEP provided superior classification performance and variable selection compared with other existing penalized methods. It is a reliable penalized method in binary logistic regression.

Spss Programs For Addressing Two Forms Of Power For Multiple Regression Coefficients, Christopher Aberson

Journal of Modern Applied Statistical Methods

This paper presents power analysis tools for multiple regression. The first takes input of correlations between variables and sample size and outputs power for multiple predictors. The second addresses power to detect significant effects for all of the predictors in the model. Both employ user-friendly SPSS Custom Dialogs.

Are Per-Family Type I Error Rates Relevant In Social And Behavioral Science?, Andrew V. Frane

Journal of Modern Applied Statistical Methods

The familywise Type I error rate is a familiar concept in hypothesis testing, whereas the per‑family Type I error rate is rarely addressed. This article uses Monte Carlo simulations and graphics to make a case for the relevance of the per‑family Type I error rate in research practice and pedagogy.

Per Family Error Rates: A Response, James F. Troendle, Keshia-Lee Martin, Vance W. Berger

Journal of Modern Applied Statistical Methods

As the authors note, the familywise error rate (FWER) is used rather often, whereas the per-family error rate (PFER) is not. Is this as it should be? It would seem that no universal answer is possible, as context determines which is more appropriate in any given application. In the general scenario of testing the benefit of an intervention, one might ideally want an error rate that aligns with the decision for benefit. In most cases the FWER does this pretty well, while allowing one to identify those endpoints for which benefit exists. The PFER does not seem to have any …

Maximum Likelihood Estimation Of The Kumaraswamy Exponential Distribution With Applications, K. A. Adepoju, O. I. Chukwu

Journal of Modern Applied Statistical Methods

The Kumaraswamy exponential distribution, a generalization of the exponential, is developed as a model for problems in environmental studies, survival analysis and reliability. The estimation of parameters is approached by maximum likelihood and the observed information matrix is derived. The proposed models are applied to three real data sets.

Test For The Equality Of Partial Correlation Coefficients For Two Populations, Madhusudan Bhandary, Arjun K. Gupta

Journal of Modern Applied Statistical Methods

A likelihood ratio test for the equality of two partial correlation coefficients based on two independent multinormal samples has been derived. The large sample Z-test for the same problem has also been discussed. The power analysis of the two tests is obtained. It has been found that the approximate likelihood ratio (ALR) test showed consistently better results than Z -test in terms of power. The size of the ALR test is slightly more than the alpha level. The ALR test is recommended strongly for use in practice.

Comparison Of Model Fit Indices Used In Structural Equation Modeling Under Multivariate Normality, Sengul Cangur, Ilker Ercan

Journal of Modern Applied Statistical Methods

The purpose of this study is to investigate the impact of estimation techniques and sample sizes on model fit indices in structural equation models constructed according to the number of exogenous latent variables under multivariate normality. The performances of fit indices are compared by considering effects of related factors. The Ratio Chi-square Test Statistic to Degree of Freedom, Root Mean Square Error of Approximation, and Comparative Fit Index are the least affected indices by estimation technique and sample size under multivariate normality, especially with large sample size.

Method Of Estimation In The Presence Of Non-Response And Measurement Errors Simultaneously, Rajesh Singh Singh, Prayas Sharma

Journal of Modern Applied Statistical Methods

The problem of estimating the finite population mean of in simple random sampling in the presence of non-response and response error was considered. The estimators use auxiliary information to improve efficiency, assuming non–response and measurement error are present in both the study and auxiliary variables. A class of estimators was proposed and its properties studied in the simultaneous presence of non-response and response errors. It was shown that the proposed class of estimators is more efficient than the usual unbiased estimator, ratio and product estimators under non-response and response error together. A numerical study was carried out to compare its …

Pseudo-Random Number Generators For Vector Processors And Multicore Processors, Agner Fog

Journal of Modern Applied Statistical Methods

Large scale Monte Carlo applications need a good pseudo-random number generator capable of utilizing both the vector processing capabilities and multiprocessing capabilities of modern computers in order to get the maximum performance. The requirements for such a generator are discussed. New ways of avoiding overlapping subsequences by combining two generators are proposed. Some fundamental philosophical problems in proving independence of random streams are discussed. Remedies for hitherto ignored quantization errors are offered. An open source C++ implementation is provided for a generator that meets these needs.

Estimating The Accuracy Of Automated Classification Systems Using Only Expert Ratings That Are Less Accurate Than The System, Paul E. Lehner

Journal of Modern Applied Statistical Methods

A method is presented to estimate the accuracy of an automated classification system based only on expert ratings on test cases, where the system may be substantially more accurate than the raters. In this method an estimate of overall rater accuracy is derived from the level of inter-rater agreement, Bayesian updating based on estimated rater accuracy is applied to estimate a ground truth probability for each classification on each test case, and then overall system accuracy is estimated by comparing the relative frequency that the system agrees with the most probable classification at different probability levels. A simulation analysis provides …

Modeling Probability Of Causal And Random Impacts, Stan Lipovetsky, Igor Mandel

Journal of Modern Applied Statistical Methods

The method of the estimation of the probability of an event occurring under the influence of the causal and random effects is considered. Epistemological differences from the traditional approaches to causality are discussed, and a new model of the statistical estimation of the parameters of each effect is proposed. The simple and effective algorithms of the model parameters estimation are presented, and numerical simulations are performed. A practical marketing example is analyzed. The results support the validity of the estimation procedure and open the perspective for the application of the method for various decision making problems, where different causes can …