R INTAMAP
From Intamap
R INTAMAP Setup
The project involves a number of partners that are all going to work on different parts of one common program package. The only feasible way to address this problem is that each group in the project implement their solutions as modules that can be attached to a main program, with a protocol for the parameters and state variables that will go in and out of these modules. A skeleton of the R package INTAMAP will soon be available for implementation of the modules. As this package develops, it will provide more accurate information about each single module than this document.
The second figure below gives a suggestion of the modularisation of the software. The figure is created in Inkscape, which runs under both Windows and Linux (www.inkscape.org). Send an email to j.skoien@geo.uu.nl if you would like a copy to be able to make changes, or send your comments to the same address. One small thing to notice is that you should make sure that the button for scaling stroke with objects should be turned off. If not, the weight of the lines scales when you change the size of an object.
Inkscape - how to turn off automatic scaling of stroke
The suggested model for structuring the INTAMAP software
For some of the functions, e.g. for spatialPredict, there are several possible methods for the solution of this task. Different methods can then be implemented as instances of this function according to the class of the first argument passed. If a user wants Bayesian kriging to be used, the modules
- estimateParameters.BayesianKriging
- spatialPredict.BayesianKriging
will be called. The prediction grid/points/lines will be associated with the class BayesianKriging in the preprocessing module, and a call to estimateParameters or spatialPredict will automatically call the proper module for this type of estimation.
The figure suggests that some of the modules are optional, implementations of them are not, and should not be necessary for the full model to deliver a result. They will mainly alter the variables in the software, not create new variables or new properties of the variables. If they create new variables of new properties, these are optional variables for the later functions. There are four functions that are not optional, and where at least one instance has to be available for testing purposes. These are:
- preProcess
- estimateParameters
- spatialPredict
- postProcess
The Help-file of the R package INTAMAP will contain information about input and output variables of each module on a later stage. These wiki pages are at the moment a more updated source for this information.
All the functions have been assigned an identification code in the figure, to make it easier to describe the use of the module. The letter refers to which part of the program the module belongs to, the number is a running number. The program has been divided into three parts, Pre-processing (PX), Interpolation KX) and Post processing (RX), where X is the number of the module.
As we have decided to use R as the computation tool in this project, all communication between modules will take place in R. All file names refer to files under the intamap project developed at the SVN repository:
http://intamap.svn.sourceforge.net/
The code is found either under R or under tests.
(For comparison only:)
