1,721,059 research outputs found
The influence of misspecification of the heteroscedasticity on multilevel regression parameter and standard error estimates
No full textLike in ordinary regression models, in multilevel analysis, homoscedasticity of the residual variances is an assumption that is mostly unchecked. However, in experimental research, the residual variance component at level two may differ in the experimental and the control condition, leading to heteroscedastic second level variances. Using a simulation study, the consequences of ignoring second level heteroscedasticity on the estimation of the fixed and random parameters and their standard errors was investigated. It was found that the standard error of the second level variance is underestimated, but that the estimated fixed parameters of the independent variables, the first level variance and their standard errors are mostly unbiased
Sample Size Determination for Bayesian Informative Hypothesis Testing
No full textThis dissertation discusses sample size determination to obtain the desired Bayes factor if the researchers use the Bayes factor to evaluate the null, unconstrained, and informative hypotheses. The informative hypothesis can express the specific expectation of the researchers through (in)equality constraints among parameters of interest in a statistical model. The evidence in favor of one hypothesis compared to another can be quantified by the Bayes factor. If the Bayes factor cannot reach a convincing value in a sample of a particular size, the study would produce an inconclusive result. Thus, Bayesian statisticians may be interested in the determination of sample sizes to obtain the desired Bayes factor. This dissertation develops an R package SSDbain to help applied researchers to plan the sample size if they use the Bayes factor to evaluate hypotheses. This package can be used to calculate the sample size for null, unconstrained, and informative hypotheses for a two-sample t-test, one-way ANOVA, and multiple linear regression. The sample size is determined such that the probability that the Bayes factor exceeds a pre-specified threshold reaches a pre-specified value. With the tool provided, the researchers can easily plan their sample size before data collection
Power analysis of longitudinal studies with piecewise linear growth and attrition
Get PDFIn longitudinal research, the development of some outcome variable(s) over time (or age) is studied. Such relations are not necessarily smooth, and piecewise growth models may be used to account for differential growth rates before and after a turning point in time. Such models have been well developed, but the literature on power analysis for these models is scarce. This study investigates the power needed to detect differential growth for linear–linear piecewise growth models in further detail while taking into account the possibility of attrition. Attrition is modeled using the Weibull survival function, which allows for increasing, decreasing or constant attrition across time. Furthermore, this work takes into account the realistic situation where subjects do not necessarily have the same turning point. A multilevel mixed model is used to model the relation between time and outcome, and to derive the relation between sample size and power. The required sample size to achieve a desired power is smallest when the turning points are located halfway through the study and when all subjects have the same turning point. Attrition has a diminishing effect on power, especially when the probability of attrition is largest at the beginning of the study. An example on alcohol use during middle and high school shows how to perform a power analysis. The methodology has been implemented in a Shiny app to facilitate power calculations for future studies. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.3758/s13428-022-01791-x
Optimal designs for group randomized trials and group administered treatments with outcomes at the subject and group level
Get PDFWith group randomized trials complete groups of subject are randomized to treatment conditions. Such grouping also occurs in individually randomized trials where treatment is administered in groups. Outcomes may be measured at the level of the subject, but also at the level of the group. The optimal design determines the number of groups and the number of subjects per group in the intervention and control conditions. It is found by taking a budgetary constraint into account, where costs are associated with implementing the intervention and control, and with taking measurements on subject and groups. The optimal design is found such that the effect of treatment is estimated with highest efficiency, and the total costs do not exceed the budget that is available. The design that is optimal for the outcome at the subject level is not necessarily optimal for the outcome at the group level. Multiple-objective optimal designs consider both outcomes simultaneously. Their aim is to find a design that has high efficiencies for both outcome measures. An Internet application for finding the multiple-objective optimal design is demonstrated on the basis of an example from smoking prevention in primary education, and another example on consultation time in primary care
Bayesian evaluation of informative hypotheses in cluster-randomized trials
No full textResearchers often have informative hypotheses in mind when comparing means across treatment groups, such as H1 : μA < μB < μC and H2 : μB < μA < μC, and want to compare these hypotheses to each other directly. This can be done by means of Bayesian inference. This article discusses the disadvantages of the frequentist approach to null hypothesis testing and the advantages of the Bayesian approach. It demonstrates how to use the Bayesian approach to hypothesis testing in the setting of cluster-randomized trials. The data from a school-based smoking prevention intervention with four treatment groups are used to illustrate the Bayesian approach. The main advantage of the Bayesian approach is that it provides a degree of evidence from the collected data in favor of an informative hypothesis. Furthermore, a simulation study was conducted to investigate how Bayes factors behave with cluster-randomized trials. The results from the simulation study showed that the Bayes factor increases with increasing number of clusters, cluster size, and effect size, and decreases with increasing intraclass correlation coefficient. The effect of the number of clusters is stronger than the effect of cluster size. With a small number of clusters, the effect of increasing cluster size may be weak, especially when the intraclass correlation coefficient is large. In conclusion, the study showed that the Bayes factor is affected by sample size and intraclass correlation similarly to how these parameters affect statistical power in the frequentist approach of null hypothesis significance testing. Bayesian evaluation may be used as an alternative to null hypotheses testing
Randomization of Clusters Versus Randomization of Persons Within Clusters: Which Is Preferable?
No full textThe effect of missing data on design efficiency in repeated cross-sectional multi-period two-arm parallel cluster randomized trials
Get PDFThe reduced efficiency of the cluster randomized trial design may be compensated by implementing a multi-period design. The trial then becomes longitudinal, with a risk of intermittently missing observations and dropout. This paper studies the effect of missing data on design efficiency in trials where the periods are the days of the week and clusters are followed for at least one week. The multilevel model with a decaying correlation structure is used to relate outcome to period and treatment condition. The variance of the treatment effect estimator is used to measure efficiency. When there is no data loss, efficiency increases with increasing number of subjects per day and number of weeks. Different weekly measurement schemes are used to evaluate the impact of planned missing data designs: the loss of efficiency due to measuring on fewer days is largest for few subjects per day and few weeks. Dropout is modeled by the Weibull survival function. The loss of efficiency due to dropout increases when more clusters drop out during the course of the trial, especially if the risk of dropout is largest at the beginning of the trial. The largest loss is observed for few subjects per day and a large number of weeks. An example of the effect of waiting room environments in reducing stress in dental care shows how different design options can be compared. An R Shiny app allows researchers to interactively explore various design options and to choose the best design for their trial
Optimal placebo-treatment comparisons in trials with an incomplete within-subject design and heterogeneous costs and variances
Get PDFThe aim of a clinical trial is to compare placebo to one or more treatments. The within-subject design is known to be more efficient than the between-subject design. However, in some trials that implement a within-subject design it is not possible to evaluate the placebo and all treatments within each subject. The design then becomes an incomplete within-subject design. An important question is how many subjects should be allocated to each combination of placebo and treatments. This paper studies optimal allocations of subjects in trials with a placebo and two treatments under heterogenous costs and variances. Two optimality criteria that consider the placebo-treatment contrasts simultaneously are considered, and the design is derived under a budgetary constraint. More subjects are allocated to those combinations with higher variances and lower costs. The optimal allocation is compared to the uniform allocation, which allocates equal number of subjects to each placebo and treatment combination, and to the complete within-subject design, where placebo and all treatments are available in each subject. The methodology is illustrated on the basis of an example on consultation time in primary care. A Shiny app is available to facilitate use of the methodology
Optimal allocations for two treatment comparisons within the proportional odds cumulative logits model
Get PDFThis paper studies optimal treatment allocations for two treatment comparisons when the outcome is ordinal and analyzed by a proportional odds cumulative logits model. The variance of the treatment effect estimator is used as optimality criterion. The optimal design is sought so that this variance is minimal for a given total sample size or a given budget, meaning that the power for the test on treatment effect is maximal, or it is sought so that a required power level is achieved at a minimal total sample size or budget. Results are presented for three, five and seven ordered response categories, three treatment effect sizes and a skewed, bell-shaped or polarized distribution of the response probabilities. The optimal proportion subjects in the intervention condition decreases with the number of response categories and the costs for the intervention relative to those for the control. The relation between the optimal proportion and effect size depends on the distribution of the response probabilities. The widely used balanced design is not always the most efficient; its efficiency as compared to the optimal design decreases with increasing cost ratio. The optimal design is highly robust to misspecification of the response probabilities and treatment effect size. The optimal design methodology is illustrated using two pharmaceutical examples. A Shiny app is available to find the optimal treatment allocation, to evaluate the efficiency of the balanced design and to study the relation between budget or sample size and power
- …
