Chapter 8 Tips for Users
PoissonERM
was firstly designed as an internal tool and is developed into a package-like version for external use. Although there are 3 main functions for completing the analysis, there are hundreds of functions behind them and none of them is supposed to be used by user directly. On the other hand, ModelPoisson()
and PredictionPoisson()
return values to the global environment and the modeling results are saved by saving the global environment into an RData object. Unlike most of the R packages, PoissonERM
uses the global environment as a painting canvas to conduct most of the operations.
Below is several suggestions for users to avoid errors while using PoissonERM
- Make sure the control script is error-free by running the options row by row.
- Always clean the global environment via rm(list = ls(all = TRUE)) before running
ModelPoisson()
. This function sources information from the control script, so it will be affected by anything that was not provided in control script but remains in the global environment from the previous run. - Similarly, remember to clean the global environment before running
PredictionPoisson()
if the last operation is notModelPoisson()
.PredictionPoisson()
sources information from the control script and loads essential modeling results from the saved RData. - The modeling results and the results with additional prediction can be saved into two different RData files. However, while running
ReportPoisson()
with the modeling results fromModelPoisson()
, make sure there is no “Prediction” folder under each endpoint’s folder. It is recommended to re-run theModelPoisson()
to clean the folders if the prediction results are not needed. - One project, one folder. The data sets and the control scripts are stored in the same directory as all analysis results. Using one folder for multiple projects may bring confusions.
- The directory path must be an absolute path. After running
ModelPoisson()
, the working directory will be the providedpathRunType
. IfpathRunType
was a relative path, it will not work inPredictionPoisson()
. It is recommended to always use an absolute path, or always usepathRunType = getwd()
(also the default value ofpathRunType
) inPredictionPoisson()
orReportPoisson()
after runningModelPoisson()
.