Statistical Methods Clause Samples

Statistical Methods. Two statistical methods were used to perform the GWAS, namely weighted single-step GBLUP (WssGBLUP; ▇▇▇▇ et al. 2012) and ▇▇▇▇▇▇ (▇▇▇▇▇▇ et al. 2011). The model adopted for WssGBLUP was: y=1µ+Zaa+e, where y is the vector of phenotypes, µ is the overall mean, a is the vector of additive genetic effects, 1 is a vector of ones, Za is an incidence matrix relating the phenotypes to a, and e is the vector of residuals. The covariance between a and e was assumed to be zero and their variances were considered to be Hσ 2 and Iσ 2, respectively, where σ 2 and σ 2 are the direct additive and residual variance, respectively, H is the matrix which combines pedigree and genomic information (▇▇▇▇▇▇▇ et al. 2010), and I is an identity matrix. The SNP effects (û) were calculated as in Stranden & ▇▇▇▇▇▇▇ (2009): û=DP’[PDP’]-1ag, where D is a diagonal matrix that contains the weights for the SNPs, P is a matrix relating genotypes of each locus (coded as 0, 1 or 2 according to the number of copies of allele B) and ag is a vector with the estimated breeding values of genotyped animals. D, â and û were iteratively recomputed over three iterations. In the first iteration (w1), the diagonal elements of D (di) were assumed to be 1 (i.e., the same weight for all markers). For the subsequent iterations (w2 and w3), di was calculated as: di=ûi2pi(1-pi), where ûi is the allele substitution effect of the ith marker, estimated from the previous iteration, and pi is the allele frequency of the second allele of the ith marker. The WssGBLUP was adopted using two sets of data, one considering all available phenotypic information (SI; n=45,000) and another considering phenotypes just from genotyped animals (SII; n=2,000). The three different weights for the SNPs (w1 to w3) and the two sets of data (SI and SII) resulted in six different solutions for the SNP effects obtained under the WssGBLUP method. BayesC was applied under the model:

Statistical Methods. A randomization analysis was conducted to check the comparability of the different conditions at baseline. This was done by chi-square statistics for categorical and dichotomous variables, while t-tests were used for continuous variables. An attrition analysis was conducted to see whether there were differences in baseline scores between the participants who remained in the study and those who withdrew at post-test. This was done by analyses of variance and chi-square. Finally, to check the effectiveness of the tailored letters, logistic regression analyses were conducted with benzodiazepine cessation at post-test as the dependent variable (‘0’- did not quit and ‘1’ – did quit) and condition as the independent variable. All comparisons between the intervention conditions were adjusted for age, gender and benzodiazepine dose (in diazepam equivalents).

Statistical Methods. 12.1.1 Comparisons of Interest [***]

Statistical Methods. All subjects who are randomized, take one or more doses of test material and have at least one post treatment efficacy measurement will be included in the analysis. Comparisons between treatments will be assessed using an *** with factors of ***, *** and *** and with *** for percent weight loss, and *** for percentage of subjects with at least 5% weight loss. A step-down multiple comparison procedure will be used to compare each dose group with ***. That is, comparison with *** will start at the ***. If the statistical test is significant at *** for both co-primary endpoints, then the test will proceed to the *** also at the ***. If the statistical comparison is not significant at the ***, then the statistical test will be stopped and the *** will not be tested. If both dose groups are significantly better than ***, then the two active dose groups will be compared. A *** of difference in response rate between treatment groups will be derived. The *** for subjects who discontinue treatment prior to completion of the study, ***. INTERNAL PROTOCOL APPROVAL 2 PRINCIPAL INVESTIGATOR SIGNATURE 3 PROTOCOL SYNOPSIS 4

Statistical Methods. Primary Analysis The primary analysis will be based on a mixed model repeated measures (MMRM) analysis to estimate the difference in change from baseline to Week 12 in mean average pain score in patients receiving ricolinostat versus those receiving placebo. The outcome will include the weekly mean change from baseline in the daily average pain scores computed each week during Week 1 through Week 12. The baseline average pain score will be computed as the mean of the average daily pain scores during the 7 days prior to randomization. The following covariates will be included in the model: baseline value of the pain score, visit, treatment group, use of concomitant medication for painful DPN (yes vs. no), a treatment group by visit interaction term, and a baseline pain score by visit interaction term. The model will first be fit with an unstructured covariance matrix. If this model does not converge, then additional structures will be considered as outlined in the protocol and Statistical Analysis Plan (SAP).

Statistical Methods. The physical addresses of all 242 Title X publicly-funded family planning clinic sites in Georgia for the year 2006 were geocoded using ArcGis (ArcMapVersion 10). Next, we used the American Community Survey data files to characterize each zip code in Georgia with regard to key population-level characteristics (e.g., proportion of households below the federal poverty line, and with characteristics such as race/ethnicity and age). The NSFG race- and age-specific estimates were used to estimate the proportion of the base population “at risk” of unintended pregnancy in each census tract and HUD tract-zip code crosswalk file were used to aggregate the population at risk in each zip code area. We then utilized Georgia’s pregnancy termination file to calculate the number pregnancy terminations occurring for each of the state’s zip codes for 2006. For each ZCTA, an abortion rate was calculated based upon the total number of terminated pregnancies over the total number of women of reproductive age (15-44 years) at risk of unintended pregnancy residing in the geographic area. The covariates were variables that were both included in our dataset and in past studies shown to be associated with unintended pregnancies. These variables included the woman’s age category (15-17, 18-19, 20-24, 25-29, 30- 34, and 35-44), ethnicity and race (non-Hispanic White, non-Hispanic Black, Hispanic, and other), percentage poverty per zip code (0-10%, 10-20%, 20-30%, and 30+%), geographical characteristic (metropolitan/urban, micropolitan/large rural, and small/isolated rural town), and distance in meters away from the closest Title X family planning clinic (0-4000, 4000-7000, 7000-11000, 11000-15000, and 15000+ meters). Distance categories were based on the quintile distribution of abortion rates occurring from the ITOP file. The sum of all abortions per variable was calculated using Statistical Analysis System (SAS Institute, Inc., Cary, NC). The abortion rate per variable was analyzed using a Poisson model (SAS Proc Genmod) utilizing robust standard errors to account for multiple observations within each zip code area. The study was approved by the Emory University Institutional Review Board [8].

Statistical Methods. The microbiome composition for each group of the different demographical ar- eas was assumed to follow a Dirichlet – multinomial distribution with 6 cate- 74 Chapter 4 – The effect of gut-microbiome on immune response gories which represents the 6 most abundant phyla (Actinobacteria, Bacteroidetes, Firmicutes, Proteobacteria, unclassified bacteria and pooled). The difference in the microbiome composition between groups was tested using the likelihood ratio test statistic with 6 degrees of freedom. ˘ All parameters of interest were described as means or frequency ( standard deviation). Prevalence rates were calculated and compared using the ▇▇▇▇▇▇▇ ▇▇▇- square test, while the Student t-test was used to compare continuous variables. To study the relationship between cytokines and microbiome over the two time points, a linear mixed effect regression model was fitted with helminth sta- tus and treatment as covariates. All models have been adjusted with age and sex, however, since both covariates were not significantly associated with the cytokine responses in any model, they are not included in the final analysis. The correla- tion between observations from the same subjects was modelled by including a subject-specific random effect. The microbiome was included in the model ei- ther as a bacterial proportion or by the ▇▇▇▇▇▇▇ diversity index. The cytokine responses were log10-transformed (log10(concentration + 1)) to obtain normally distributed variables. First, we analyzed the main effect of bacterial proportion and diversity on cytokine responses. Second, to allow for different effect sizes of bacterial proportion or diversity on cytokine responses in helminth-positive versus -negative subjects, an interaction term between bacterial categories and infection was included in the model. The p-value for this interaction term in- dicated the statistical evidence for different effect sizes in helminth-positive or –negative groups. To allow the estimation of the treatment effect on the relationship between bacterial proportion and cytokine responses, the randomized controlled trial de- sign was used. Since the sample size is too small, we only stratified based on randomization arm. Hence, the effect of treatment cannot be distinguished from the effect of helminth infection. Therefore, we explored the relationship between cytokines and microbiome after anthelminthic treatment. A linear mixed effect model was fitted with bacterial proportion or diversity, and treatme...

Statistical Methods. The objective of this Phase II study is to evaluate the safety of IM administration of VM202 in subjects with moderate or high-risk CLI (▇▇▇▇▇▇▇▇▇▇ Clinical Severity Score equal to 4 or 5) who are poor or non- candidates for surgical or percutaneous revascularization; and, to evaluate potential bioactivity of IM administration of VM202 in subjects with CLI, when compared with placebo, on rest pain (as assessed by frequency of rest pain, pain medication use history, sleeping history, and intensity of rest pain) or leg ulcer healing (as assessed by ulcer surface area, time to complete healing), perfusion (MRA), hemodynamic assessment (ABI & TBI), tissue oxygenation (TcPO2) and the incidence and extent of lower leg amputation or other surgical interventions.

Statistical Methods. NIH-PA Author Manuscript χ2 tests were used to assess differences in the distributions of categorical phenotypes between the random sample of sibling used in this study and the cohort of concordant siblings from which the random sample was taken. A 2-sample t test for independent samples was used to compare age between the 2 groups. Interrater reliability was assessed using the kappa (κ) statistic. We used the guidelines proposed by ▇▇▇▇▇▇ and ▇▇▇▇ to interpret the level of agreement: κ>0.80, excellent agreement; κ=0.60 to 0.80, substantial agreement; κ=0.40 to 0.60, moderate agreement; and κ<0.40, poor to fair agreement (17). The following stroke subtypes were used: 1) large vessel, 2) cardioembolism, 3) small vessel, 4) other determined cause, and 5) undetermined cause. A sixth category—no stroke/missing—was included to account for instances in which reviewers did not provide a diagnosis. Reasons for missing data were not systematically collected. Separate estimates for κ were obtained by treating the no stroke/missing category as missing data using the algorithm written by ▇▇▇▇ et al (18). These analyses were repeated on the subset of the participants who self-reported having a stroke. ▇▇▇▇▇▇▇ et al. Page 4 NIH-PA Author Manuscript The randomly sampled siblings in this study were screened for participation in SWISS between March 31, 2000, and November 6, 2003. The 30 siblings included were from 30 separate pedigrees. The probands of the siblings were enrolled at centers in 8 different states in the United States. Among the 30 siblings evaluated in this study, head computed tomography (CT) was performed on 80% (24/30) and MRI was performed on 60% (18/30). A cervical arterial imaging study (carotid ultrasonography, magnetic resonance angiography, or digital subtraction angiography) was performed on 97% (29/30). Intracranial arterial study (magnetic resonance angiography, digital subtraction angiography, CT angiography, or transcranial Doppler ultrasonography) was performed on 100% (30/30). Electrocardiography was performed on 73% (22/30). Cardiac ultrasonography (transthoracic or transesophageal) was performed on 67% (20/30). NIH-PA Author Manuscript Subject characteristics for the siblings included in this study are summarized in Table 1. The majority of subjects were male (67%) and white (83%). Ages ranged from 46 to 88 years, with a median of 70 years. Seventy-three percent smoked, 70% had hypertension, 27% had diabetes mellitus, 33% had chronic fibril...