Analyst Application ANOVA
Analysis of Variance
The ANOVA tasks in the Analyst Application build in complexity. The One-Way ANOVA offers the most basic specification, and the Factorial ANOVA extends the possible model to multiple classification variables. A Linear Models task provides a more comprehensive set of tools for modeling any number of response variables by both quantitative and class independent variables. Finally, you can use the Repeated Measures task to model correlated responses using a wide range of covariance structures, and you can specify a model with both fixed and random effects using the Mixed Models task.
One-Way Analysis of Variance
In the One-Way ANOVA task, you can perform a one-way analysis of variance (ANOVA). This analysis technique is used for experimental data in which there are a continuous response variable and a single independent classification variable. In the One-Way ANOVA task, you can produce a variety of tests, request means comparisons and plots, and specify By group variables. This task is appropriate if you have a single classification variable and the data are balanced.
Factorial ANOVA
In the Factorial ANOVA task, you can perform a factorial analysis of variance. This analysis technique is used for experimental data in which there are a continuous response variable and one or more independent classification variables. This analysis is appropriate if you have multiple classification variables or a single classification variable with unbalanced data (not all levels have the same number of data points).
Linear Models
In the Linear Models task, you can fit general linear models with the method of least squares. This analysis technique is used for experimental data in which there are a continuous response variable and one or more independent classification variables as well as one or more independent quantitative variables. In the linear models task, you can specify a sophisticated model, produce a variety of tests, and request means comparisons and least squares means. You can also request plots displaying interactions, predicted values, and a number of diagnostic statistics.
You can use the Analyst Application to perform repeated measures analysis. Repeated measures refer to multiple measurements on an experimental unit, such as right- and left-eye visual acuity, or measurements made over a period of time, such as blood pressure measured once a week for a month. Questions you want to answer in a repeated measures analysis may be whether treatment influences the response, whether time, the repeated factor, influences the response, and whether there is a treatment by time interaction.
Using the repeated measures task, you can specify a model, repeated effect, subject effect, and covariance structure for the measurements on an individual subject. Your model can incorporate both quantitative and classification variables for main effects, interactions, polynomial terms, and nested effects. The same covariance structure is used for all subjects, and you can select from a wide range of structures, including unstructured, autoregressive(1), and compound symmetry. You can request hypothesis tests, produce least squares means, compute descriptive statistics, and plot means, predicted values, and residuals. The task uses the mixed models approach for analyzing repeated measures.
View an example of a Repeated Measures analysis using the Analyst Application.
In the Mixed Models task, you can fit linear models that incorporate both fixed and random effects. With random effects, the levels in the factor that are used in the study represent a random sample of a larger set of possible levels. The distribution of the possible levels of the factor has a mean and a variance. Often, the random effect is that of a blocking factor, such as blocks in a split plot design. Using the Mixed Models task, you can specify fixed and random effects that include interactions, polynomials, and nesting. You can produce tests concerning fixed effects and variance components, request a specific estimation method, and compute least squares means and pairwise differences. In addition, you can output predicted values as well as plot means, predictions, and residuals.
Statistics and Operations Research Home Page | Analyst Application
