E. Niclas Jonsson and Mats O. Karlsson
Department of Pharmacy, University of Uppsala
One of the keys to rapid and successful data analysis is the use of informative graphics, both as a way to reveal structures in the data and as a diagnostic tool. With increasing complexity of the data, as in population analysis, the need for such graphical methods becomes even more important although, with increasing complexity the number of different plots can become quite large. A majority of the plots will be similar from analysis to analysis and the task of producing these plots can be automated. As the need for graphical inspection of a fit varies during the course of an analysis, a graphical tool also has to be flexible with respect to plots displayed.
Xpose is a collection of Splus functions for making various plots to aid data set checkout, goodness of fit analysis and model building. Data set checkout plots include plots of data set items vs ID number and pairwise scatter plots of covariates and goodness of fit/model building tools include basic goodness of fit plots, some more unusual goodness of fit plots (e.g. auto-correlation plots), individual plots of predicted and observed data vs the independent variable, plots of covariates vs parameters, generalized additive modeling (GAM) to find candidate covariate relationships, tree based model analysis and GAM analysis of bootstrapped data sets.
The user interface is a simple system of menus and the plots can either be sent directly to a printer or be displayed on screen, optionally in multiple windows to facilitate comparison between two or more runs.
The input to Xpose is one or more of of six standard NONMEM table files. These have a predefined format of which one has a fixed format, i.e. include the same things from run to run and from data set to data set, while the other five vary depending on the number of covariates and parameters in the analysis. In addition to the six table files, some checkout plots use the data file given to NONMEM, the only constraint being the placement of the ID column.
Since Xpose is a collection of Splus functions it is quite straightforward to customize the plots, making it easy to include Xpose plots in reports and similar documents.
Xpose has been developed for Unix (Splus v. 3.2 or later) and Windows (Splus v 3.3). It can be obtained without cost at http://xpose.sourceforge.net/.
Reference: PAGE 5 () Abstr 563 [www.page-meeting.org/?abstract=563]
Poster: oral presentation