Fixing Scree Graph

Author: rolandojduartem

FTSExample/FTSExample.ipynb

@@ -1,5 +1,5 @@
 """
-Note: V 0.9.1 Originally, filling data methods was developed by Eric Alfaro and Javier Soley in SCILAB
+Note: V 0.9.2 Originally, filling data methods was developed by Eric Alfaro and Javier Soley in SCILAB
       Python version was developed by Rolando Duarte and Erick Rivera
       Centro de Investigaciones Geofísicas (CIGEFI)
       Universidad de Costa Rica (UCR)
@@ -154,25 +154,33 @@ def checkPrincipalComponents(self):
         upperError: int
             Maximum value to choose principal components 
         """
+        #Scalating to get the best performance using PCA
         scale = StandardScaler()
         dfMeanScaled = scale.fit_transform(self.dfMean)
         pca = PCA(n_components = self.dfColumns, copy = True, svd_solver = "full", random_state = 0)
-        pca.fit(dfMeanScaled)
+        vectorsPCA = pca.fit_transform(dfMeanScaled)
         explainedVariance = pca.explained_variance_
-        errorExplainedVarience = explainedVariance * sqrt(2 / self.dfRows)
+        errorExplainedVarience = []
+
+        #Calculating error bars
+        for index in arange(0, len(explainedVariance)):
+            dfComponents = DataFrame({"Original": vectorsPCA[:, index]})
+            dfComponents["Shift"] = dfComponents.Original.shift(1)
+            corr = dfComponents.corr().iloc[0, 1]
+            nEffective = self.dfRows * (1 - corr**2) / (1 + corr**2)
+            errorExplainedVarience.append(explainedVariance[index] * sqrt(2 / nEffective))
         components = arange(1, len(explainedVariance) + 1)
         upperError = len(explainedVariance) - 1
 
-        figure(figsize = (20, 20))
+        #Plotting eigenvalues and principal components
         errorbar(components, explainedVariance, 
-                                    yerr=errorExplainedVarience, fmt="o", color="#9b6dff", 
-                                    ecolor="black", capsize=6,
+                                    yerr=errorExplainedVarience, fmt="D", color="green", 
+                                    ecolor="red", capsize=10,
                                     )
         title("Explained variance vs. principal components")
         xlabel("Principal components")
         ylabel("Explained variance")
         show()
-
         return upperError
 
     def PCAMethod(self, components=1, tol=1e-1, itermax=10, valueMin=0.0):
@@ -247,7 +255,8 @@ def checkPrincipalComponents(self):
         upperError: int
             Maximum value to choose principal components 
         """
-        return PrincipalComponentAnalysis(self.df).checkPrincipalComponents()
+        upperError = PrincipalComponentAnalysis(self.df).checkPrincipalComponents()
+        return upperError
 
     def FullMethod(self, lags=1, components=1, tol=1e-1, itermax=10, valueMin=0.0):
         """

FillingTimeSeries/FillingMethods.py

@@ -1,5 +1,5 @@
 """
-Note: V 0.9.1 Originally, filling data methods was developed by Eric Alfaro and Javier Soley in SCILAB
+Note: V 0.9.2 Originally, filling data methods was developed by Eric Alfaro and Javier Soley in SCILAB
       Python version was developed by Rolando Duarte and Erick Rivera
       Centro de Investigaciones Geofísicas (CIGEFI)
       Universidad de Costa Rica (UCR)

FillingTimeSeries/PreprocessingFillingMethods.py

@@ -1,4 +1,4 @@
-# Filling Time Series (v.0.9.1)
+# Filling Time Series (v.0.9.2)
 
 ## Filling missing values in geophysical time series
 
@@ -10,9 +10,7 @@
 Filling Time Series is a Python package to help the users to work with geophysical time series by filling missing values in their data. Filling Time Series was developed at the Centro de Investigaciones Geofísicas (CIGEFI), Universidad de Costa Rica (UCR).
 
 ## Last updates
-- Added new method: ComponentsAutoregression applies autoregression method, then, the principal components method to fill the time series.
-- Small fixes in the Scree graph.
-- Renaming some classes, please, check documentation.
+- Fixing the error bar calculation.
 
 ## Documentation
 Documentation will be available soon.
@@ -44,6 +42,10 @@ Bug reports can be submitted to the issue tracker at:
 
 [https://github.com/rolandojduartem/FillingTimeSeries/issues](https://github.com/rolandojduartem/FillingTimeSeries/issues)
 
+## References
+- Alfaro, E., & Soley, J. (2009). Descripción de dos métodos de rellenado de datos ausentes en series de tiempo metereológicos. Revista de matemática: Teoría y Aplicaciones, 16, 60 - 75.
+- Ureña, P., Alfaro, E., & Soley, J. (2016). Propuestas metodológicas para el rellenado de datos ausentes en series de tiempo geofísicas. Guía Práctica de uso. Universidad de Costa Rica.
+
 ## License
 
 MIT License

README.md

@@ -7,15 +7,15 @@
 setup(
   name = 'FillingTimeSeries',    
   packages = ['FillingTimeSeries'], 
-  version = '0.9.1',   
+  version = '0.9.2',   
   license='MIT', # License from here: https://help.github.com/articles/licensing-a-repository
   description = 'Filling Time series: Package to fill missing values in geophysical time series in Python', # Short description
   long_description = long_description,
   long_description_content_type = "text/markdown",
   author = 'Rolando Jesus Duarte Mejias, Erick Rivera Fernandez',    
   author_email = 'rolando.duartemejias@ucr.ac.cr',
   url = 'https://github.com/rolandojduartem/FillingTimeSeries', # github or website
-  download_url = 'https://github.com/rolandojduartem/FillingTimeSeries/archive/refs/tags/v_0_9_1.tar.gz', # Download file
+  download_url = 'https://github.com/rolandojduartem/FillingTimeSeries/archive/refs/tags/v_0_9_2.tar.gz', # Download file
   keywords = ['Time Series', 'Missing values', 'Metereology', "Geophysics", "Metereological"],
   setup_requires=['wheel'],
   install_requires=[       # I get to this in a second