Data Analyses. All analyses were stratified by the severity of hemophilia, and often by age category as well. As the clinical characteristics of hemophilia A and hemophilia B do not differ, we present combined results for hemophilia A and B. Data on the treatment modality, the number of bleeding episodes, the use of hospital facilities, and absence from school or work referred to the year that preceded the questionnaire surveys (2000). Children were defined as patients younger than 16, adolescents as patients between 16 and 25 and adults as patients older than 25 years. The use of prophylaxis refers to patients who received prophylaxis as their main treatment modality, excluding patients who received a combination of on demand treatment and prophylaxis during risk periods. Absence from school was calculated only for that part of the population that followed a full-time education. Absence from work was calculated for patients aged 16 to 65 who had a paid job (full-time or part-time). The inactivity ratio was calculated as the ratio of inactivity in the study population and inactivity in Dutch men. Patients that did not have a full-time or part-time paid job were defined as inactive. Descriptive statistics for age, the use of hospital facilities, absence from work and employment were compared to national figures for the general male population that were provided by the Central Bureau of Statistics Netherlands Statline database20. Self reported measures on joint impairment were obtained for a series of 16 joints which are, the neck, the left and right shoulder, the back, the left and right elbow, the left and right wrist, the left and right hand and fingers, the left and right hip, the left and right knee and the left and right ankle. The possible scores were 0 (no impairment), 1 (some impairment without daily problems), 2 (some impairment with daily problems), and a maximum of 3 (severe impairment with complete loss of function). From scores of the 16 separate joints a joint score was calculated with a minimum score of 0 and a maximum score of 48 points. As joint impairment was reported most frequently in the ankles, elbow and knees these were analyzed separately. Results Response and patient characteristics Response was 70% in 2001, compared to 84% in 197219, 70% in 197821, 81% in 198522 and 78% in 199218. One hundred and ninety eight patients participated in all 5 surveys. Table 1 shows the characteristics of participants in each of the 5 surveys. The mean age of partici...
Data Analyses. Preliminary analyses Because of the group format of the intervention, the participants in the IC could not be considered independent observations. To assess the amount of variance attributable to group differences a random coefficient regression model (RCRM) was used to estimate the intra-class correlation in the IC over the four assessment times. Preliminary analyses included checks for normality and the computation of descriptive statistics. All variables were distributed acceptably close to normal. They were described computing frequencies, means and standard deviations. Analysis of variance (ANOVA) and χ2-tests were used to compare the following groups on baseline characteristics: (a) the randomized participants and those who refused to be randomized (but had no objections to being interviewed), (b) participants randomized to IC and to WLC, (c) those dropping out of the intervention and those who complied with the course, and (d) the participants who left the study after the posttreatment assessment and those who completed both FU assessments. Pre- and posttreatment - controlled data The effects of the intervention were assessed on the completers sample by using a 2 x 2 x 2 split-plot design, using the presence of MDD (Depression) and IC as between- subject variables and time as within-subject variable (Depression x Condition x Time). Main and interaction effects were tested using the mulitvariate criterion of Xxxxx’ lambda (Λ). The analyses were repeated (a) with comorbid anxiety disorder as a third between-subjects factor and (b) with the CES-D on the intention-to-treat (ITT) sample, which included the subjects who dropped out of the IC. A last-value-carried-forward procedure was used to provide data for missing values that occurred because of dropout. To assess the clinical significance of change on an outcome measure in a clinical population Xxxxxxxx and Xxxxx (1991) proposed two criteria: (a) the change should move the individual outside the range of the dysfunctional population (referred to as change in status) and (b) the change should be statistically reliable and exceed the measurement error (referred to as reliable change). We did not assess clinical diagnoses at posttreatment, but used the score on the CES-D as an indication of functional status: those with a score ≥ 16, the recommended cut point (Xxxxxxx et al., 1997; Xxxxxxx and Xxxx, 1986) were considered to be dysfunctional. A change in status was defined as a change from a pretreatment ...
Data Analyses. To compare the incidence of plural agreement over dialects, pronoun types, and head-noun types, analyses of variance were performed on the proportions of validPlural responses for each participant andeach item in each cell of the experimental design. The proportions were calculated relative to all valid Plural and Singular re- sponses in each condition for each type of preamble. Prior to analysis the proportions were arcsin transformed(Xxxxx 1976). Analyses were performedwith both participants and items as random factors using the min F′ statistic (Xxxxx 1973). Unless otherwise indicated, effects reported as significant were associated with probabilities less than or equal to 0.05, and the corresponding test statistics are summarized in Appendix D. Type of preamble presentation (read-aloud or reproduction) was treated as a separate factor in the analyses. Because the major findings were similar regardless of presentation mode, we omit differences associated with presentation from the results and discussion.
Data Analyses. It was first examined whether those with missing values on one or more variables of interest (n=453) differed from those without missing values (n=1,363) on SES (by means of t test) and gender (by means of Pearson chi-square test). Next, it was investigated whether control variables should be included in the main statistical analyses as covariates (i.e., whether cannabis users differed from nonusers on these variables) using t tests or GLM univariate analysis of variance for SES, parental psychopathology, externalizing behaviour problems, and using Pearson chi-square analysis for gender, alcohol, and tobacco use. It was then tested whether these variables were related to social skills using Pearson correlation for SES, parental psychopathology, externalizing behaviour, and using t-tests or GLM univariate analysis of variance for gender, alcohol, and tobacco use. Logistic regression analysis was performed to examine the impact of social skills on whether or not cannabis was used during adolescence. Using multinomial regression analysis, onset of cannabis use was predicted by social skills using the three-category variable (a) no use (reference group), (b) early onset, and (c) late onset. Next, multinomial regression analyses were used to predict frequency of cannabis use at T3 from social skills, using a five-category cannabis variable as the dependent variable:
Data Analyses. All smallmouth bass captured within each of the sub-reaches will be enumerated in 2020-2021 similar to that during former years (2004 – 2019). Total numbers of smallmouth bass, largemouth bass and walleye collected and catch/effort (fish/hr) will be also determined for each sub-reach per sampling pass. Length data will be recorded for 2020-2021 similar to that during former years (2004 – 2019) to determine the size structure of smallmouth bass removed. Data analyses similar to that employed between 2004 and 2019 will be used to analyze the 2020-2021 field data.
Data Analyses. Data were entered into an electronic database and translated into English using the original English survey as a reference. Descriptive statistics of all characteristics and factors of interest by intention to intervene for each outcome of interest are summarized in Table 1. Initially, age and academic year were considered in the analytical models, however, age was removed due to evidence of multicollinearity using the variance decomposition proportion value (VDP1 > 0.7). A two-step multiple imputation method of missing values was performed to allow maximum utilization of available data. This method was recommended for imputing arbitrary patterned continuous variables and models containing mixed covariates (Xxxxx & Kosten, 2017). Multivariate Poisson regressions with robust variance analyses were performed using PROC GENMOD with a log link function to produce prevalence ratio (PRs), 95% confidence intervals (CI), standard errors (SEs), and p-values. Four models were examined with the intention to intervene in each binary outcome. Models 1.a & 1.b represent the regression results for the outcome of interest where friends are perpetrators, Models 2.a & 2.b represent the regression results for the outcome of interest where strangers are perpetrators (Table 2). Models 1.b. & 2.b. adjusted for 4 bystander variables based on the 5-step situational bystander model. Clustering by main faculties ( Humanities, Health, and Sciences) was controlled for in the imputation and regression analyses models. All statistical analysis and imputations were performed using SAS 9.3.
Data Analyses. First the correlations between the explicit and implicit object relations instruments were examined. The validity of the ♙TGR scales was examined by (a) the multidimensional scaling method (MDS), (b) testing proportions of expected stronger correlations between scales, (c) testing differences in correlations and (d) examination of individual significant correlations between scales. MDS is a statistical technique that uses proximity data ̶distances between objects̶ and transforms these into a visual representation in which the estimated position of each scale is based on the strength of all correlations between the scales. Compared to the often used “eyeball” inspection of the correlation matrix to look for patterns of associations, this visual representation has the advantage that it is relatively easy to see, for example, whether the implicit God representation scales are more strongly associated with implicit than with explicit object-relation measures. MDS searches for the optimal positioning of points in which the distances be- tween these points match best with all the proximities between the objects, and pro- vides coordinates and a geometrical representation of these positions. This is done by minimalizing the stress; the difference between estimated distances and raw proximity data. We applied this method with the SPSS-procedure PROXSC♙L (Busing, Commandeur, Heiser, Bandilla, & Xxxxxxxx, 1997). We let PROXSC♙L assign the location of the scales of ♙TGR and QGR in a two-dimensional space, based on the correlation matrix of the observed correlations between all scales as measures of prox- imity. Thereto we transformed the values of the correlations into distances (𝛿) with the following formula: 𝛿 = √2 ∗ (1 − |𝑟|) (1) There are some rules of thumb to establish the goodness of fit of the found solu- tion, but these, according to Xxxx, Xxxxxxx, and Xxxx (2012), are not very reliable because there are many aspects that need to be considered when judging stress. In this study we used the Normalized Raw Stress-value (NRS). ♙n NRS value of 0 means absolute fit, but the ideal NRS value is .02, according to XxXxxxx (2011). Because we have a theoretical model to compare the found solution to, we reported the various stress-values but did not reject, based on these subjective criteria for bad fit, solutions. We compared solutions that treated distances as ordinal and were based on a classical Xxxxxxxxx start configuration with those with multiple random start...
Data Analyses. Scale reliabilities, descriptive statistics, intercorrelations, and multiple linear regression models were computed using SAS software, Version 9.4 (SAS Institute Inc.) or IBM SPSS Statistics, Version 26 (IBM Corp.). Path models were assessed using Structural Equation Modeling (SEM) with Mplus Version 8.6 (Xxxxxx & Xxxxxx, 1998-2017). Multi-item scales were specified in such models as latent constructs, each measured, relying on the accepted approach of parcelling (Little et al., 2013), with three indicators calculated as random thirds of the scale items. Participants’ sociodemographic and medical characteristics were modelled as observed variables. In models involving multiple measurements of the same construct, variance resulting from specific measurement occurrences was accounted for by correlating all the measurement errors of same indicators across time points (Xxxxx & Xxx, 1996). To assure weak factorial invariance, factor loadings were constrained for equality across measurement waves. As there were missing values in the data, and the data deviated from normality, we used the Mplus MLR estimator that allows for maximum likelihood estimation with robust standard errors and chi-square calculation in presence of missing values (Xxxxxx & Xxxxx, 2003). Following recommendations of Xx and Xxxxxxx (1999), we report two fit indexes: Xxxxxx-Xxxxx Index (TLI) and Comparative Fit Index (CFI), and two indexes of misfit: Root Mean-Square Error of Approximation (RMSEA) and Standardised Root Mean-Square Residual (SRMR) are reported. TLI and CFI close to or above 0.95, combined with RMSEA below 0.06 and SRMR below 0.08, are considered indicative of acceptable fit.
Data Analyses. Testing proportions of stronger correlations between scales. We com- pared the (absolute) strength of correlations of implicit versus explicit God represen- tation scales with the explicit personality pathology scales by computing six propor- tions per group: each proportion represents the number of comparisons with stronger associations of the five personality scales with a specific QGR scale than with a specific ♙TGR scale, divided by the total number of compared associations per QGR scale (25). The sixth proportion (per group) was the sum of the proportions per QGR scale, divided by the total number of all comparisons (150). The significances of proportions of stronger associations were tested by a binomial test, performed in EXCEL with the formula BINOM.DIST (number_s, trials, probability_s, cumulative). For the first argument (number of successes) we filled in the number of comparisons with stronger associations for the same-method than for the mixed-method combination, for the second (trials) we filled in the total number of comparisons, for the third argument (the probability of success) we filled in .5, and for the fourth we filled in ʻTrueʼ, which yields the cumulative probability. If the proportion found was higher than 0.5, we used the formula 1-BINOM.DIST; if it was lower than 0.5, we used the formula BI- NOM.DIST. Because this test assumes that the comparisons are independent, the correlations with the ♙GC subscales were left out of these analyses.
Data Analyses. The nutrition and women’s empowerment field notes were analyzed for the purposes of this thesis. The student leads on nutrition and women’s empowerment combined both the young and older women field notes to form 11 site briefs for each of the topics. Therefore, 22 site briefs were read, ‘memo-ed’, and coded using MAXQDA10 Qualitative Analysis Software and two codebooks were developed. Codes were developed inductively and deductively with a thematic analysis approach[6]. Thereafter, codebooks were shared between the student leads on nutrition and women’s empowerment and coding strategies reviewed for each domain. Coding and codebooks were revised and fieldnotes were re-coded as needed. Inter-coder reliability was assessed and the discrepancies were addressed in 5 of the site briefs. Triangulation of the final analyses was conducted to further validate the findings.
