This study examines the effects of unequal variance within groups (but not among groups) on the F, Kruskal-Wallis, and Normal Scores tests. The study simulates the unequal variance problem for three distributions, two configurations of means, three levels of means, two different numbers of groups, three sample sizes per group, and equal or unequal variance within group. The distribution affected the performance of all tests, but the F test was affected most. Violation of the equal variance assumption harmed the F test severely, although not as much as non-normality. The Kruskal-Wallis and Normal Scores tests behaved almost identically.
The lack of consensus regarding outcome measures and trial design issues in the idiopathic inflammatory myopathies (IIM) is inhibiting the conduct and interpretation of clinical trials. To begin to address these problems, a multispecialty group of over 70 adult and pediatric neurologists, rheumatologists, rehabilitation medicine physicians, statisticians, and patient support group leaders, called the International Myositis Outcome Assessment Collaborative Study Group (IMACS), is engaged in developing consensus on the assessment of disease activity and damage for myositis clinical trials. As part of this ongoing international effort, members of this group met in November 2001 at a work-shop entitled "Defining Clinical Improvement in Adult and Juvenile Myositis." A goal of the work-shop was to review current data on the validity and responsiveness of the recently published proposed preliminary core set measures for disease outcome assessment in clinical trials for myositis and to define the degree of change in each core set measure that is clinically meaningful. Despite differences in the clinical presentations, natural history and responses to therapy between adult onset and juvenile onset myositis, expert specialists in these diseases came to a consensus that the amount of improvement that is clinically meaningful in each core measure is the same for adult and juvenile myositis. For the domains of muscle strength and physical function, a minimum of 15% improvement is clinically significant, whereas for the physician and patient global assessments, as well as the extramuscular assessment, a minimum of 20% improvement is considered clinically meaningful, and for serum levels of muscle associated enzymes, at least 30% improvement is needed to be clinically important. This workshop is the first of several planned to develop multidisciplinary, international consensus on the conduct and reporting of IIM clinical trials.
Using data from 3 independent studies, to quantify the interobserver reliability of semi-quantitative skin scoring methods (the original and the modified Rodnan skin thickness scores) used to assess the degree and extent of cutaneous thickening in systemic sclerosis (SSc).Interobserver variability of the original Rodnan skin thickness score method (cutaneous thickness assessed in 26 body surface areas using a 0-4 scale) was evaluated in one study. The modified Rodnan method (cutaneous thickness assessed in 17 body surface areas using a 0-3 scale) was evaluated in 2 studies. In all 3 studies, each patient's skin thickness was assessed by 6 or 7 observers in a blinded fashion.The overall within patient standard deviations were not statistically different in all 3 studies (5.4, 4.6 and 4.6) irrespective of the overall mean skin thickness scores (26.6, 18.3 and 17.7). With the original Rodnan technique, the within patient standard deviation tended to be higher in patients with higher skin thickness scores. In the 2 studies which used the modified technique, no significant differences in within patient standard deviation were noted between high and low skin thickness scores.Three independent studies demonstrate that the Rodnan skin thickness scoring techniques are reproducible among different observers (the within patient standard deviation being consistently about 5 units). Our data provide valuable information needed for sample size calculations for SSc trials in which skin thickness score is an outcome variable.
This prospective study examined the frequency and severity of respiratory illnesses in survivors of preterm birth compared with those in full-term infants. Although preterm infants did not demonstrate an inherent risk of subsequent respiratory illness when compared with full-term infants, earlier and more severe lower respiratory illnesses were observed among survivors of idiopathic respiratory distress syndrome (RDS). Infants who survived RDS but who developed residual lung disease had a greater risk of both more frequent and more severe subsequent lower respiratory illnesses than did RDS survivors who did not have persistent roentgenographic changes. We suggest that the risk of increased respiratory illness in these infants was a consequence of residual pulmonary abnormalities apparent on the chest roentgenogram at discharge from the nursery. Agents associated with respiratory illnesses were similar in all groups of study patients.
Abstract This is used to compare mean differences between treatments when the observations have been obtained in pairs. The difference between the paired values is assumed to be normally distributed, and the null hypothesis that the expectation is zero is tested by Student's t test. The robustness properties are discussed, as is the asymptotic relative efficiency of nonparametric alternatives.