Two-Stage Testing for Genome-Wide Gene-Environment Interactions
Ørnulf Borgan, Norman E. Breslow, Nilanjan Chatterjee, Mitchell H. Gail, Alastair Scott, Christopher J. Wild in Handbook of Statistical Methods for Case-Control Studies, 2018
When a two-stage testing procedure for genome-wide GE is used, the distributions of test statistics for those genetic variants that pass the first-stage filtering are conditional on the corresponding filtering statistics being greater than a pre-specified threshold. It is therefore different from their marginal distributions. Multiple-testing adjustment, for example a Bonferroni correction, has to account for the conditional distributions induced by such screening. However, if the screening statistic and the testing statistic are independent, the screening step can be ignored and the multiple-testing adjustment is only needed for genetic variants being tested in the second stage. This seemingly intuitive result is formally proved for the family-wise error rate in Theorem 1 in Dai et al. (2012).
Changes in Mobility of Children with Cerebral Palsy Over Time and Across Environmental Settings
Robert J. Palisano in Movement Sciences: Transfer of Knowledge into Pediatric Therapy Practice, 2012
Post hoc comparisons for the setting main effect (data collapsed over time) indicated that there were significant differences between all three settings (Figure 2), based on the ranks of mobility methods from 1 (highest) to 9 (lowest). The unadjusted p-values are presented and using the Bonferroni correction would require a p < 0.0167 to be significant. In addition, 98.3% Confidence Intervals (CI) are presented to be consistent with a Bonferroni correction for the number of comparisons. At home, children used higher-ranked mobility methods (M= 4.01) than at school (M = 4.68): mean difference = .672 (98.3% CI = .217 to 1.127), SE = .185, t = 3.64, p ≤ 0.001. At school, children used higher-ranked mobility methods (M= 4.68) than in the outdoors/community (M= 5.90): mean difference = 1.22 (98.3% CI = .679 to 1.762), SE = .220, t = 5.55, p ≤ .0001. The mean difference between home (M = 4.01) and outdoors/community (M = 5.90) was significant: mean difference = 1.89 (98.3% CI = 1.30 to 2.485), SE = .241, t = 7.87, p ≤ 0.0001. The comparison of the three settings, collapsed over time, was replicated with a Friedman rank ANOVA, followed by Bonferroni-corrected Wilcoxon tests, due to the partially ordinal nature of the data. The results (overall p ≤ 0.0001 and all three settings differed from each other at p ≤ 0.001) were similar to the results from the parametric ANOVA.
Traditional Methods for Analysis of Longitudinal and Clustered Data
Gueorguieva Ralitza in Statistical Methods in Psychiatry and Related Fields, 2017
When observations are made at multiple intermediate time points, some studies still report results from a t-test or ANOVA performed separately at each time point. However, if a 5% significance level test is used for each test, then the probability of type I error becomes considerably greater than 5% and this may lead to falsely declaring groups to be different when differences do not exist. On the other hand, if a procedure for correcting for multiple testing is used (e.g., Bonferroni correction), then this leads to loss of power and increase in the probability of type II error (not detecting treatment differences when they do exist). Performing separate tests at each time point also does not allow for direct comparison between treatment groups over time. This approach is not considered in this chapter as it rarely makes sense for longitudinal data.
Demographics, self-efficacy, benefits/barriers, stages of adopting pap testing among Korean American women
Published in Journal of Obstetrics and Gynaecology, 2019
Wei-Chen Tung, Michelle Granner, Minggen Lu, Jeeyun Sohn
Proportions and counts were applied to describe demographic variables and distributions of TTM stages. Averages and standard deviations (SD) were used to describe self-efficacy, perceived benefits and barriers scores. Chi-square tests were used to assess the associations between TTM stages and categorical variables. The average self-efficacy was obtained by computing the average across seven items from the self-efficacy scale. The averages of three items from the perceived benefits subscale and 12 items from the perceived barriers subscale were computed to obtain the means of the perceived benefits and barriers. Linear regression models were applied to compare the means of self-efficacy and perceived benefits and barriers between TTM stages, after adjusting for potential confounding variables (i.e. age, marital status, education, abnormal Pap smear). Post hoc multiple comparisons were performed to detect the differences between specific pairs of changes in outcomes. The level of significance for multiple comparisons was adjusted by Bonferroni correction. The results were regarded as significant if p values < =.05.
Mimicking chronic glaucoma over 6 months with a single intracameral injection of dexamethasone/fibronectin-loaded PLGA microspheres
Published in Drug Delivery, 2022
Alba Aragón-Navas, María J. Rodrigo, David Garcia-Herranz, Teresa Martinez, Manuel Subias, Silvia Mendez, Jesús Ruberte, Judit Pampalona, Irene Bravo-Osuna, Julian Garcia-Feijoo, Luis E. Pablo, Elena Garcia-Martin, Rocío Herrero-Vanrell
Data were recorded in an Excel database and statistical analysis was performed using SPSS software version 20.0 (SPSS Inc., Chicago, IL). The Kolmogorov-Smirnov test was used to assess sample distribution. Given the parametric distribution of the data, Student’s t-test was used to evaluate the differences between eyes, and a paired Student’s t-test was used to compare the changes recorded in each eye over the study period. All values were expressed as means ± standard deviations. Values of p < 0.05 (expressed as *) were considered to indicate statistical significance. The Bonferroni correction for multiple comparisons was also calculated to avoid a high false-positive rate. The level of significance for each variable was established according to Bonferroni calculations (expressed as #).
Vitamin D receptor gene polymorphisms and risk of polycystic ovary syndrome in South Indian women
Published in Gynecological Endocrinology, 2018
Swapna Siddamalla, Tumu Venkat Reddy, Suresh Govatati, Nagendram Erram, Mamata Deenadayal, Sisinthy Shivaji, Manjula Bhanoori
To analyze the combined effect of VDR SNPs on PCOS development, the haplotype frequencies for multiple loci and the standardized disequilibrium coefficient (D′) for pair-wise LD were estimated (Table 3). The LD between three loci was not much different between cases and controls (Figure 2). Our data suggests that the BsmI A, ApaI A and TaqI T, is the most common haplotype in South Indian women. Hence, the relative risk of each haplotype was calculated by using this as reference (Table 3). Bonferroni correction was used to adjust the significance level. Since we have eight haplotypes, the Bonferroni correction should be 0.05/8 = 0.0062. Therefore, a p values <.0062 was considered significant. Our results showed significant association between haplotypes G/A/C (p= .0031) and A/C/C (p= .0004), while remaining haplotypes were not indicative for disease risk.
Related Knowledge Centers
- Null Hypothesis
- Power of A Test
- Statistical Hypothesis Testing
- Type I & Type II Errors
- Multiple Comparisons Problem
- Type I & Type II Errors
- P-Value
- Family-Wise Error Rate
- Holm–Bonferroni Method
- Expected Value
- Power of A Test