Home
Welcome to the evolving documentation for the PhenStat package. As opposed to the twice-yearly bioconductor releases, PhenStat undergoes many incremental changes and bug fixes. This wiki is intended to bridge the documentation gap between the time of the last official bioconductor release of PhenStat and any updates since.
is a package that provides statistical methods for the identification
of abnormal phenotypes with an emphasize on high-throughput data-flows.
The package contains dataset checks and cleaning in preparation for the
analysis, five statistical frameworks for the phenodeviants identification and
additional functions that help to decide the correct method for analysis.
Download Package
Download UserGuide
View Code
Bioconductor
- Install dependencies
install.packages("methods")
install.packages("car")
install.packages("nlme")
install.packages("nortest")
install.packages("vcd")
install.packages("MASS")
install.packages("logistf")
install.packages("lme4")
- Download the latest version of the package:
https://github.com/mpi2/stats_working_group/blob/master/PhenStat/PhenStatPackage/PhenStat_2.6.0.tar.gz?raw=true - Install
install.packages("path_to_downloaded_file",repos=NULL,type="source")
-
NEW FUNCTIONALITY
- PhenList object stores unfiltered dataset for the possible future usage in RR and FE frameworks.
- testDataset function: argument's "useUnfiltered" with default value set to FALSE. Can be used to process dataset in RR and FE frameworks with unfiltered by genotype/sex combination records (see p.21 of User Guide).
-
NEW FUNCTIONALITY
- testDataset function: argument's "transformationValue" default value is set to FALSE.
- New graph functions: boxplotSexGenotypeResult, scatterplotSexGenotypeBatchResult and scatterGenotypeWeightResult. The same functionality as for original functions but PhenTestResult object is an argument.
- New adjustment functionality: getColumnBatchAdjusted and getColumnWeightBatchAdjusted functions returns PhenList column values adjusted for batch or for batch and weight accordingly.
- Columns with adjusted values in PhenTestResult object:
- "variable_name" - transformed or raw values used in the modelling process;
- "variable_name_original" - raw values for variable;
- "variable_name_batch_adjusted" - raw values adjusted for batch effect;
- "variable_name_batch_weight_adjusted" - raw values adjusted for batch and weight effect.
- Graph functions for adjusted values: boxpolotSexGenotypeBatchAdjusted - creates a box plot split by sex and genotype after accounting for batch, boxplotSexGenotypeWeightBatchAdjusted - creates a box plot split by sex and genotype after accounting for batch and weight.
-
NEW FUNCTIONALITY
- Ability to determine and apply the optimal transformation of continuous data. This is based around the Box-Cox transformation.
- An additional statistical framework is implemented for the categorical data – Biased Reduction Logistic Regression.
- New visualisation function: boxplotSexGenotypeBatchAdjusted that allows data to be visualise after batch has been accounted for.
-
REFACTORING
- Major architectural changes in object structure. PhenList and PhenTestResult are now S4 classes and new methods are provided for the classes.
- The PhenTestResult class contains the dataset used for the analysis and associated calculated values (e.g. transformed, batch adjusted)
-
BUG FIX
A bug in the summaryOutput with the FE framework has been fixed to ensure numbers are appropriate reported. -
COMPATIBILITY ISSUES
The output from vectorOutput function has an additional element for transformation values.