Python Forum

Hello!

Im trying to replace invalid values ( x< -3 and x >12) with 'nan's in a pandas data structure . The first two columns consist of ids and names respectively, and should not be modified. I want to make a general code for data with an unknown amount of column values, I know that the first two columns are ids and names but don't know the amount columns of number data that I will have to deal with.

My question is: How would I be able to modify data, replacing invalid values with nans for the third column and onwards (range(2,Columns)) ?

Using below mentioned code replaces all names and ids with nan, but does not do anything to the number data.

Data[(Data>=12) & (Data<=-3] = np.nan 

Using below mentioned code replaces all rows, from 2 to 6, with nan values.

Data[(Data.loc[range(2,Columns)] >=12) & (Data.loc[range(2,Columns)]<=-3)] = np.nan

I have uploaded a csv file..

Kind regards

The entirety of the code;

import numpy as np
import chardet
import pandas as pd

# Open and read csv file 
with open('testfilex1w.csv', 'rb') as f:
    
    #detect encoding of csv file, assigning encoding to -Result-
    Result = chardet.detect(f.read())
    
# Use panda to read csv file relative to above detected encoding    
Data = pd.read_csv('testfilex1w.csv', encoding=Result['encoding'], header= None)

#Display duplicates (student ids) found in pd.dataframe -Data-
print(Data[Data.duplicated([0], keep=False)])

# Drop duplicated rows based on IDs[0] and Names[1], IDS and Names are defined
# below

Data = Data.drop_duplicates([0], keep='last')
Data = Data.drop_duplicates([1], keep='last')



# Compute number of rows and columns of original data
Columns = len(Data.columns)
Rows    = len(Data.index)

# Create a selection of columns to group data
Ids = Data.iloc[:, 0]
Names = Data.iloc[:, 1]

# The amount of grades columns is unknown in the assignment, I must therefore
# create code that will work with x amount of columns
Grades = Data.iloc[:, range(2,Columns)]

# Compute number of rows and columns of dupliate ridden grades
Gradecols = len(Grades.columns)
Graderows =len(Grades.index)

Elements = (Gradecols * Graderows)

# Create a "memory" array of ones for indexing, in order to remove rows from
# original grades data by modifying this array with zeros based on below for/ if loop.
#Valid = pd.DataFrame(np.ones((len(Grades.index),len(Grades.columns))))


# Create variable to keep track of rows, where invalid data might exist
Rowcount=0
Colcount = 0
Gradecount = 0

# Create matrix array of grades... because I cant figure out how to iterate over
# a pandas dataframe :(
Gradesnp = Grades.values
Valid=np.ones(Elements)
# For loop with if statements in order to find occurences of values out of 
# range and modify the memory array with zeros so that I can remove out of range
# values from orginal grades data.


for row in Gradesnp:
    for item in row:
        if(item < -3.0 or item > 12.0):
            print("Invalid grade found! Grade was {} in Line {}.".format(item, Rowcount))
            #modify -Valid- array per index, with zero if the if statement is satisfied.
            Gradecount+=1
            Valid[Rowcount]=0
            
    Rowcount+=1
    

        

# Through the use of Boolean indexing, the data array is modified by the 
# memory vector(-Valid-) so that values corresponding to 0 are excluded and values 
#corresponding to 1 are included.
#Validgrades=Gradesnp[Valid==1,:]


#Data[(Data>=12) & (Data<=-3] = np.nan
     
#Data[(Data.iloc[:, range(2,Columns)]>=12) & (Data.iloc[:, range(2,Columns)]<=-3)] = np.nan

#Data.loc[range(2,Columns)] = np.nan  

#Data.loc[range(2,Columns)] 

#Data[(Data.loc[range(2,Columns)] >=12) & (Data.loc[range(2,Columns)]<=-3)] = np.nan

You guys are slow.

#Replace invalid grades with Nans

Data.iloc[:,2:] = Data.iloc[:,2:].where((Data < 12) & (Data > -3))