dataset
"""
import spectrochempy as scp
############################################################
# Load a dataset
dataset = scp.read_omnic("irdata/nh4y-activation.spg")
print(dataset)
dataset.plot_stack()
##############################################################
# Create a PCA object
pca = scp.PCA(dataset, centered=False)
##############################################################
# Reduce the dataset to a lower dimensionality (number of
# components is automatically determined)
S, LT = pca.reduce(n_pc=0.99)
print(LT)
###############################################################
# Finally, display the results graphically
# ScreePlot
_ = pca.screeplot()
########################################################################################################################
import spectrochempy as scp
import sys
############################################################
# Upload a dataset form a distant server
try:
dataset = scp.download_iris()
except (IOError, OSError):
print("Could not load The `IRIS` dataset. Finishing here.")
sys.exit(0)
##############################################################
# Create a PCA object
pca = scp.PCA(dataset, centered=True)
##############################################################
# Reduce the data to a lower dimensionality. Here, the number of
# components is automatically determined using `n_pc="auto"`. As
# indicated by the dimension of LT, 4 PC are found.
S, LT = pca.reduce(n_pc="auto")
print(LT)
###############################################################
# The figures of merit (explained and cumulative variance) confirm that
# these 4 PC's explain 100% of the variance:
#
pca.printev()
# %% [markdown] # ### Guessing the concentration profile with PCA + EFA # # Generally, in MCR ALS, the initial guess cannot be obtained independently of the experimental data 'x'. # In such a case, one has to rely on 'X' to obtained (i) the number of pure species and (ii) their initial # concentrations or spectral profiles. The number of of pure species can be assessed by carrying out a PCA on the # data while the concentrations or spectral profiles can be estimated using procedures such EFA of SIMPLISMA. # The following will illustrate the use of PCA followed by EFA # # #### Use of PCA to assess the number of pure species # # Let's first analyse our dataset using PCA and plot a screeplot: # %% pca = scp.PCA(X) pca.printev(n_pc=10) _ = pca.screeplot(n_pc=8) # %% [markdown] # The number of significant PC's is clearly larger or equal to 2. It is, however, difficult tto determine whether # it should be set to 3 or 4... Let's look at the score and loading matrices: # # %% S, LT = pca.reduce(n_pc=8) _ = S.T.plot() _ = LT.plot() # %% [markdown] # Examination of the scores and loadings indicate that the 4th component has structured, non random scores and loadings.
# %% [markdown] # ### Guessing the concentration profile with PCA + EFA # # Generally, in MCR ALS, the initial guess cannot be obtained independently of the experimental data 'x'. # In such a case, one has to rely on 'X' to obtained (i) the number of pure species and (ii) their initial # concentrations or spectral profiles. The number of pure species can be assessed by carrying out a PCA on the # data while the concentrations or spectral profiles can be estimated using procedures such EFA of SIMPLISMA. # The following will illustrate the use of PCA followed by EFA # # #### Use of PCA to assess the number of pure species # # Let's first analyse our dataset using PCA and plot a screeplot: # %% pca = scp.PCA(X) pca.printev(n_pc=10) _ = pca.screeplot(n_pc=8) # %% [markdown] # The number of significant PC's is clearly larger or equal to 2. It is, however, difficult tto determine whether # it should be set to 3 or 4... Let's look at the score and loading matrices: # # %% S, LT = pca.reduce(n_pc=8) _ = S.T.plot() _ = LT.plot() # %% [markdown] # Examination of the scores and loadings indicate that the 4th component has structured, nonrandom scores and loadings.