In sanchezi/phisStatR: Phenoarch Greenhouse Data Analysis

---

title: "Phenoarch platform - Cleaning procedure - Curve level - gss package" author: "I.Sanchez" date: "r format(Sys.time(), '%B %d, %Y')" output: rmarkdown::pdf_document vignette: > %\VignetteIndexEntry{Phenoarch platform - Cleaning procedure - Curve level - gss package} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8}

---

---

Objective of cleaning procedure using smoothing splines anova

Smoothing spline analysis of variance on each genotype-scenario of an experiment. Detection of outlier repetition if significant TT*Rep (thermal time by repetition) interaction using a Kullblack-Leibler projection (KL). I consider a genotype-scenario as outlier:

• biovolume: if KL > 0.05
• plantHeight: if KL > 0.05
• leafArea: if KL > 0.05

The input dataset must contain the following columns:

• experimentAlias
• genotypeAlias
• scenario
• repetition
• thermalTime (for thermal time)
• parameter of interest (biovolume, plantHeight etc...)

The five first column names are standard names extracted from the web service.

Import of data

  library(ggplot2)
  library(lubridate)
  library(tidyr)
  library(dplyr)
  library(gss)
  library(phisStatR)

  myreport<-substr(now(),1,10)

  data(plant3)
  cat("-------------- plant3 dataset ---------------\n")  
  printExperiment(datain=plant3)

  # Import data, here is a dataset in the phisStatR package, You have to import your own dataset
  # using a read.table() statement or a request to the web service
  # You can add some datamanagement statements...
  #------------------------------------------------------------------------
  # Please, add the 'Ref' and 'Genosce' columns if don't exist. 
  # 'Ref' is the concatenation of experimentAlias-Line-Position-scenario
  # 'Genosce' is the concatenation of experimentAlias-genotypeAlias-scenario
  #------------------------------------------------------------------------

  mydata<-unite(plant3,Genosce,experimentAlias,genotypeAlias,scenario,sep="-",remove=FALSE)
  mydata<-arrange(mydata,Genosce)

  # For one parameter, for example biovolume
  resbio<-fitGSS(datain=mydata,trait="biovolume",loopId="Genosce") 

Curves by genotype-scenario

Biovolume

  outlierbio<-printGSS(object=resbio[[2]],threshold = 0.05)
  klbio<-printGSS(object=resbio[[2]],threshold = NULL)

  cat("Detection of outlier curve with KL projection:\n")
  print(outlierbio)

  #------------------------------------------------
  # You can export these two datasets
  # suppress the comments
  #------------------------------------------------
  #write.table(outlierbio,paste0(myreport,"outlier_gss_biovolume.csv"),row.names = FALSE,sep="\t")
  #write.table(klbio,paste0(myreport,"KLprojection_gss_biovolume.csv"),row.names = FALSE,sep="\t")

I take a threshold of 0.05 for this example. We can take a more conservative threshold like 0.01 or 0.02 to detect more outlier curves...

  # plot of the smoothing splines by genotype-scenario  
  for(i in seq(1,length(unique(mydata[,"Genosce"])),by=12)){ 
    myvec<-seq(i,i+11,1)
    myvec<-myvec[myvec<=length(unique(mydata[,"Genosce"]))]
    print(plotGSS(datain=mydata,modelin=resbio[[1]],trait="biovolume",myvec=myvec,lgrid=50))
    cat("\n\n")
  }

Session info

  sessionInfo()

References

1. R Development Core Team (2015). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org.
2. Chong Gu (2014). Smoothing Spline ANOVA Models: R Package gss. Journal of Statistical Software, 58(5), 1-25. URL http://www.jstatsoft.org/v58/i05/.

sanchezi/phisStatR documentation built on Nov. 14, 2019, 7:10 p.m.