Lite: iNZight Lite

Model Fitting

This module is designed to fit regression models from the data. The controls on the left are divided into six sections. The Main panel on the right shows model summaries plots or code to produce the models. - Select model In the first section all models are stored. From a drop down menu a model can be selected, renamed with a text field next to it by typing in a new name and pressing the "RENAME" button. The "REMOVE" button removes a model from the list. There is no model to choose from by default. Underneath the model controls is a select bar which lets the user switch between the "Model" summary, "Plots" or the "Code History".

Model

- Choose Model settings This seetings makes it possible to select the Y variable from the data. Transform the Y variable by log, sqare root or raising it by a factor which can be specified in a text filed appearing in the same row. The model framework can be selected and whether the design is a complex survey or not. -Predictor variables The next section is for choosing predictor variables or confounding variables. - Interaction terms Simple interactions can be specified with the next section. The choices are "All" (All possible interactions), "by degree" (in a text field the degree of interaction is supplied), "by variable" (specific interaction between variables). - Transform x variables Different transformations of the x variables can be specified. - Display Formula The code which will produce the module is presented in this section.

Plots

Three different plot categories can be selected. - Factor level comparison This plot can be produced for every factor in the fitted model. It shows how each factor for a variable influenes the model. - Graphical diagnostics Six different plots can viewed. A plot of the residuals versus the fitted values, a scale location plot, The residuals versus leverage, a cooks distance plot, a normall QQ-plot and a histogram of the residuals. - Normality checks The Normal QQ-plot is repeated here and the histogram as well as a histogram of sampled data and a QQ-plot including inference.