Assessing Equivalent and Inverse Change in Genes Between Diverse Experiments

Background: It is important to identify when two exposures impact a molecular marker (e.g., a gene’s expression) in similar ways, for example, to learn that a new drug has a similar effect to an existing drug. Currently, statistically robust approaches for making comparisons of equivalence of effect sizes obtained from two independently run treatment versus control comparisons have not been developed. Results: Here, we propose two approaches for evaluating the question of equivalence between effect sizes of two independent studies: a bootstrap test of the Equivalent Change Index (ECI), which we previously developed, and performing Two One-Sided t-Tests (TOST) on the difference in log-fold changes directly. The ECI of a gene is computed by taking the ratio of the effect size estimates obtained from the two different studies, weighted by the maximum of the two p-values and giving it a sign indicating if the effects are in the same or opposite directions, whereas TOST is a test of whether the difference in log-fold changes lies outside a region of equivalence. We used a series of simulation studies to compare the two tests on the basis of sensitivity, specificity, balanced accuracy, and F1-socre. We found that TOST is not efficient for identifying equivalently changed gene expression values (F1-score = 0) because it is too conservative, while the ECI bootstrap test shows good performance (F1-score = 0.96). Furthermore, applying the ECI bootstrap test and TOST to publicly available microarray expression data from pancreatic cancer of tumor tissue and peripheral blood mononuclear cells (PBMC) showed that, while TOST was not able to identify any equivalently or inversely changed genes, the ECI bootstrap test identified genes associated with pancreatic cancer, intestinal cancer, and other disease types associated with pancreatic cancer. Conclusion: A bootstrap test of the ECI is a promising new statistical approach for determining if two diverse studies show similarity in the differential expression of genes and can help to identify genes which are similarly influenced by a specific treatment or exposure.

FSTruct: An Fst-based tool for measuring ancestry variation in inference of population structure

In model-based inference of population structure from individual-level genetic data, individuals are assigned membership coefficients in a series of statistical clusters generated by clustering algorithms. Distinct patterns of variability in membership coefficients can be produced for different groups of individuals, for example, representing different predefined populations, sampling sites, or time periods. Such variability can be difficult to capture in a single numerical value; membership coefficient vectors are multivariate and potentially incommensurable across groups, as the number of clusters over which individuals are distributed can vary among groups of interest. Further, two groups might share few clusters in common, so that membership coefficient vectors are concentrated on different clusters. We introduce a method for measuring the variability of membership coefficients of individuals in a predefined group, making use of an analogy between variability across individuals in membership coefficient vectors and variation across populations in allele frequency vectors. We show that in a model in which membership coefficient vectors in a population follow a Dirichlet distribution, the measure increases linearly with a parameter describing the variance of a specified component of the membership vector. We apply the approach, which makes use of a normalized Fst statistic, to data on inferred population structure in three example scenarios. We also introduce a bootstrap test for equivalence of two or more groups in their level of membership coefficient variability. Our methods are implemented in the R package FSTruct.

A hierarchical clustering method for random intervals based on a similarity measure

A new clustering method for random intervals that are measured in the same units over the same group of individuals is provided. It takes into account the similarity degree between the expected values of the random intervals that can be analyzed by means of a two-sample similarity bootstrap test. Thus, the expectations of each pair of random intervals are compared through that test and a p-value matrix is finally obtained. The suggested clustering algorithm considers such a matrix where each p-value can be seen at the same time as a kind of similarity between the random intervals. The algorithm is iterative and includes an objective stopping criterion that leads to statistically similar clusters that are different from each other. Some simulations to show the empirical performance of the proposal are developed and the approach is applied to two real-life situations.

Comparing Alternative Corrections for Bias in the Bias-Corrected Bootstrap Test of Mediation

Although the bias-corrected (BC) bootstrap is an often-recommended method for testing mediation due to its higher statistical power relative to other tests, it has also been found to have elevated Type I error rates with small sample sizes. Under limitations for participant recruitment, obtaining a larger sample size is not always feasible. Thus, this study examines whether using alternative corrections for bias in the BC bootstrap test of mediation for small sample sizes can achieve equal levels of statistical power without the associated increase in Type I error. A simulation study was conducted to compare Efron and Tibshirani’s original correction for bias, z 0, to six alternative corrections for bias: (a) mean, (b–e) Winsorized mean with 10%, 20%, 30%, and 40% trimming in each tail, and (f) medcouple (robust skewness measure). Most variation in Type I error (given a medium effect size of one regression slope and zero for the other slope) and power (small effect size in both regression slopes) was found with small sample sizes. Recommendations for applied researchers are made based on the results. An empirical example using data from the ATLAS drug prevention intervention study is presented to illustrate these results. Limitations and future directions are discussed.

A Bootstrap Method to Test Granger-Causality in the Frequency Domain

We propose a bootstrap test for unconditional and conditional Granger-causality spectra in the frequency domain. Our test aims to detect if the causality at a particular frequency is systematically different from zero. In particular, we consider a stochastic process derived applying independently the stationary bootstrap to the original series. At each frequency, we test the sample causality against the distribution of the median causality across frequencies estimated for that process. Via our procedure, we infer about the relationship between money stock and GDP in the Euro Area during the period 1999–2017. We point out that the money stock aggregate M1 had a significant impact on economic output at all frequencies, while the opposite relationship is significant only at low frequencies.

A Robust Bootstrap Test for Mediation Analysis

Mediation analysis is central to theory building and testing in organizational sciences. Scholars often use linear regression analysis based on normal-theory maximum likelihood estimators to test mediation. However, these estimators are very sensitive to deviations from normality assumptions, such as outliers, heavy tails, or skewness of the observed distribution. This sensitivity seriously threatens the empirical testing of theory about mediation mechanisms. To overcome this threat, we develop a robust mediation method that yields reliable results even when the data deviate from normality assumptions. We demonstrate the mechanics of our proposed method in an illustrative case, while simulation studies show that our method is both superior in estimating the effect size and more reliable in assessing its significance than the existing methods. Furthermore, we provide freely available software in R and SPSS to enhance its accessibility and adoption by empirical researchers.

The Effect of Forbush Decreases on Atmospheric Aerosols and Clouds from The PATMOS-x Satellite from 1978 to 2018

<p>The solar system is constantly changing, and it is important for us to understand how our climate and weather changes in response to the solar activity during both long-time scales (e.g. the 11-year solar cycle) and short time scales (e.g. days to weeks during For-bush Decreases (FDs)). Solar variability causes a corresponding modulation of the incident number of cosmic rays in Earth's atmosphere. Previous work by [Veretenenko and Pudovkin, 1997], [Svensmark and Friis-Christensen, 1997], [Palle Bago and Butler, 2000], [Svensmark et al., 2016], [Harrison and Ambaum, 2010], and other researchers have discussed this cause-effect relationship from an experimental and theoretical approach. Since the 1970s, global observations of the Earth's system by satellites are offering an invaluable source of information about cloud parameters.</p><p>In this study, we used the newly calibrated PATMOS-x (Pathfinder Atmospheres Extended) data set during the period from 1978 to the present. A method for capturing the connection between cosmic rays and meteorological measurements has been conducted by superposition analysis of FD events for time series (36 days) and the Monte Carlo bootstrap test to evaluate significance level of the integrated signal for 9 days after the minimum in FD. We have reviewed results, primarily about cloud emissivity (Achieved Significance Level (ASL >99%), surface brightness temperature (ASL >99%), and cloud fraction (ASL >99%). Some of the results support the proposed relationship between solar activity and temperature. This result indicates that the amount of incident cosmic rays decreases due to FDs, global average temperature increases [Friis-Christensen and Lassen, 1991], [Harrison and Ambaum, 2010]. In addition, PATMOS-x parameters of cloud probability, cloud mask, and cloud fraction, which all means cloud coverage on the Earth shows statistically significant signals following FDs. In some previous research, IR-detected cloud fraction from International Satellite Cloud Climate Project (ISCCP) and combined liquid and ice cloud fraction, effective emissivity from the Moderate Resolution Imaging Spectroradiometer (MODIS) also show connection with FDs, see [Svensmark et al., 2009], [Svensmark et al., 2016], [Marsh and Svensmark, 2000a], Todd and Kniveton [2004]. The relationship between the observed changes in cloud amount and the resulting solar forcing is discussed. On the other hand, “Cloud water content" from Special Sensor Microwave Imager (SSM/I), “Liquid water path", and “Optical thickness" from MODIS also showed as significant signals by FDs, see [Svensmark et al., 2009], [Svensmark et al., 2016], however a similar parameter about “optical thickness" and “integrated total cloud water over whole column g/m2" from PATMOS-x dataset does not have high significant signals by a bootstrap test with ASL of 77.03 and 92.51% respectively. Moreover, significant results are reported for several new cloud parameters from the PATMOS-x dataset (e.g. cloud type, brightness temperature, measurements by different wavelength 0.65, 0.86, 3.75, 11.0, and 12.0 μm and others) and Fu-Liou model is used for estimation of changed radiations in the atmosphere. An interaction between CCN and radiation has not been investigated well yet. It is necessary to still more to learn about these results for further understanding of Earth’s atmosphere.</p>

Testing the Number of Components in a Birnbaum-Saunders Mixture Model under a Random Censoring Scheme

This paper deals with testing the number of components in a Birnbaum-Saunders mixture model under randomly right censored data. We focus on two methods, one based on the modified likelihood ratio test and the other based on the shortcut of bootstrap test. Based on extensive Monte Carlo simulation studies, we evaluate and compare the performance of the proposed tests through their size and power. A power analysis provides guidance for researchers to examine the factors that affect the power of the proposed tests used in detecting the correct number of components in a Birnbaum-Saunders mixture model. Finally an example of aircraft Windshield data is used to illustrate the testing procedure.