Adding a outlier detection before the PLSR calibration. It uses a LOO…

… validation to identify the most uncertain points and remove them from the total dataset
plantphys · Nov 8, 2021 · bfb9395 · bfb9395
1 parent 5c591b5
commit bfb9395
Show file tree

Hide file tree

Showing 11 changed files with 5 additions and 0 deletions.
diff --git a/Overal_PLSR_modeling.pptx b/Overal_PLSR_modeling.pptx
diff --git a/PLSR/PLSR_models.R b/PLSR/PLSR_models.R
@@ -30,6 +30,11 @@ hist(data_curated$sqrt_Vcmax25_JB)
 ## Removing Na values to avoid any issue
 data_curated=data_curated[!is.na(data_curated$sqrt_Vcmax25),]
 
+## Using a LOO PLSR model to identify wrong data
+test_LOO=plsr(sqrt_Vcmax25~ Spectra,ncomp = 19, data = data_curated, validation = "LOO")
+res_LOO=test_LOO$validation$pred[,1,19]-data_curated$sqrt_Vcmax25
+outliers=which(abs(res_LOO)>3*sd(res_LOO))
+data_curated=data_curated[-outliers,]
 ############################
 #### PLSR models VCMAX25 ###
 ############################

diff --git a/PLSR/Validation_Albert_et_al_2018.pdf b/PLSR/Validation_Albert_et_al_2018.pdf
diff --git a/PLSR/Validation_Alldatasets 2021-11-08 sqrt_Vcmax25_JB.pdf b/PLSR/Validation_Alldatasets 2021-11-08 sqrt_Vcmax25_JB.pdf
diff --git a/PLSR/Validation_Alldatasets 2021-11-08.pdf b/PLSR/Validation_Alldatasets 2021-11-08.pdf
diff --git a/PLSR/Validation_dataset sqrt_Vcmax25_JB .jpeg b/PLSR/Validation_dataset sqrt_Vcmax25_JB .jpeg
diff --git a/PLSR/Validation_dataset sqrt_Vqmax25_JB .jpeg b/PLSR/Validation_dataset sqrt_Vqmax25_JB .jpeg
diff --git a/PLSR/Validation_random sqrt_Vcmax25_JB .jpeg b/PLSR/Validation_random sqrt_Vcmax25_JB .jpeg
diff --git a/PLSR/Validation_random sqrt_Vqmax25_JB .jpeg b/PLSR/Validation_random sqrt_Vqmax25_JB .jpeg
diff --git a/PLSR/Validation_random.jpeg b/PLSR/Validation_random.jpeg
diff --git a/PLSR/Validation_random_JB.jpeg b/PLSR/Validation_random_JB.jpeg