Remove outlier with Python

Question

I have a DataFrame which consists of 30 rows and 9 columns. I want to make a 2 sigma outlier removal.

I do it with this:

from scipy import stats
df[(np.abs(stats.zscore(df)) < 2).all(axis=1)]

But it removes the whole line if there is a outlier in a single column. I just want to get this single value deleted. How can I do this? And the first column contains the time. this should never be touched. How can I exclude this single column?

this is how the data looks like:

Trace for Mass: 60Ni    61Ni    62Ni    63Cu    64Ni    65Cu    66Zn
Resolution: High    High    High    High    High    High    High
                            
Time    Intensity   Intensity   Intensity   Intensity   Intensity   Intensity   Intensity
[sec]   [cps]   [cps]   [cps]   [cps]   [cps]   [cps]   [cps]

0.  4.246875178068876e-003  4.550645244307816e-004  8.364085806533694e-004  3.21496045216918e-003   3.215973265469074e-003  1.595904817804694e-003  1.983924303203821e-003  
1.051999807357788   4.264393821358681e-003  5.171436932869256e-004  8.292743586935103e-004  3.154967911541462e-003  3.216561861336231e-003  1.622977200895548e-003  1.874359208159149e-003  
2.102999925613403   4.27544629201293e-003   4.796394787263125e-004  8.318902109749615e-004  3.211528761312366e-003  3.147452371194959e-003  1.622740761376917e-003  1.879810937680304e-003  
3.154999971389771   4.278738517314196e-003  4.829006502404809e-004  7.972901221364737e-004  3.218628698959947e-003  3.22998408228159e-003   1.604416524060071e-003  1.938240835443139e-003  
4.206999778747559   4.211603198200464e-003  4.424861108418554e-004  8.007381693460047e-004  3.2428870908916e-003    3.166524693369865e-003  1.590821426361799e-003  1.903632888570428e-003  
5.257999897003174   4.267803858965635e-003  5.1306706154719e-004    8.309389813803136e-004  3.144200425595045e-003  3.117314074188471e-003  1.603707205504179e-003  1.815222087316215e-003  
6.309999942779541   4.182798787951469e-003  5.052632768638432e-004  7.896805764175952e-004  3.130593337118626e-003  3.10095027089119e-003   1.570251770317555e-003  1.817710697650909e-003  
7.361000061035156   4.296375438570976e-003  4.910536226816475e-004  8.9122453937307e-004    3.204192267730832e-003  3.028199542313814e-003  1.533132861368358e-003  1.788084045983851e-003  
8.413000106811523   4.335530567914248e-003  6.025235052220523e-004  8.631621603854001e-004  3.268211148679256e-003  2.987353131175041e-003  1.608435995876789e-003  1.796260941773653e-003  
9.463999748229981   4.290143493562937e-003  4.839488829020411e-004  8.525795419700444e-004  3.222533734515309e-003  3.005951410159469e-003  1.583610195666552e-003  1.700276043266058e-003  
10.51599979400635   4.287909716367722e-003  5.497571546584368e-004  9.083477198146284e-004  3.219338599592447e-003  2.950039459392428e-003  1.682562520727515e-003  1.783343963325024e-003  
11.56699943542481   4.260278772562742e-003  4.665948799811304e-004  7.738673011772335e-004  3.193542594090104e-003  2.853760728612542e-003  1.568833249621093e-003  1.736654434353113e-003  
12.61899948120117   4.26474679261446e-003   5.00720867421478e-004   8.611407829448581e-004  3.217800287529826e-003  2.865647897124291e-003  1.595077337697148e-003  1.658685388974845e-003  
13.67099952697754   4.222772549837828e-003  4.647313617169857e-004  8.633999968878925e-004  3.159464336931706e-003  2.801976399496198e-003  1.629361184313893e-003  1.673259655945003e-003  
14.72200012207031   4.23405971378088e-003   4.880253691226244e-004  8.320091292262077e-004  3.10550956055522e-003   2.766199875622988e-003  1.57923623919487e-003   1.671363832429051e-003  
15.77400016784668   4.263806156814098e-003  5.268111126497388e-004  8.335548918694258e-004  3.150589996948838e-003  2.747958991676569e-003  1.52225757483393e-003   1.638660905882716e-003  
16.82500076293945   4.173276014626026e-003  5.153965321369469e-004  7.848058012314141e-004  3.132368205115199e-003  2.736426191404462e-003  1.501098275184631e-003  1.646955031901598e-003  
17.87699890136719   4.209604579955339e-003  4.582091642078012e-004  7.977656787261367e-004  3.183129709213972e-003  2.714420203119516e-003  1.604771241545677e-003  1.606788486242294e-003  
18.92900085449219   4.214542452245951e-003  4.919854109175503e-004  8.5032032802701e-004    3.177686594426632e-003  2.588512841612101e-003  1.560558215714991e-003  1.607973361387849e-003  
19.97999954223633   4.171629901975393e-003  4.438837058842182e-004  8.449696470052004e-004  3.142070723697543e-003  2.649111207574606e-003  1.58833886962384e-003   1.547667197883129e-003  
21.0310001373291    4.234999883919954e-003  5.094563821330667e-004  8.215457201004028e-004  3.189756069332361e-003  2.645698608830571e-003  1.556538976728916e-003  1.515797688625753e-003  
22.08300018310547   4.159520845860243e-003  5.21336798556149e-004   7.7945546945557e-004    3.093914361670613e-003  2.504269825294614e-003  1.597914495505393e-003  1.550629152916372e-003  
23.13399887084961   4.095097538083792e-003  5.284418002702296e-004  8.160762954503298e-004  3.164552384987474e-003  2.605574205517769e-003  1.5143376076594e-003    1.545534702017903e-003  
24.18600082397461   4.190911073237658e-003  4.741653683595359e-004  8.253505802713335e-004  3.078178269788623e-003  2.457562601193786e-003  1.61718437448144e-003   1.502647297456861e-003  
25.23799896240234   4.155758768320084e-003  4.477270995266736e-004  8.012137841433287e-004  3.119352972134948e-003  2.549331868067384e-003  1.551455701701343e-003  1.538307638838887e-003  
26.28899955749512   4.055834375321865e-003  4.267746699042618e-004  8.247561054304242e-004  3.050019731745124e-003  2.364743268117309e-003  1.565523212775588e-003  1.418655156157911e-003  
27.34099960327148   4.160813987255096e-003  4.637996316887438e-004  8.405701955780387e-004  3.15011665225029e-003   2.621341263875365e-003  1.558548538014293e-003  1.534871873445809e-003  
28.39200019836426   4.123781807720661e-003  5.418366636149585e-004  8.308201213367283e-004  3.128936979919672e-003  2.427210099995136e-003  1.607372076250613e-003  1.475754892453551e-003  
29.44400024414063   4.185620695352554e-003  4.987408174201846e-004  7.421225891448557e-004  3.080426249653101e-003  2.371448557823896e-003  1.567532890476286e-003  1.444243011064827e-003  
30.49600028991699   4.092158749699593e-003  5.319360643625259e-004  8.368841372430325e-004  3.113200422376394e-003  2.385094529017806e-003  1.580300158821046e-003  1.433581346645951e-003  

This file is read by:

pd.options.display.float_format = '{:.4f}'.format

data = pd.read_csv(dateiname, sep='\t', names=['Time', '60Ni', '61Ni', '62Ni', '63Cu', '64Ni', '65Cu', '66Zn'], skiprows=6, nrows=30, index_col=False, dtype=float)

Answer 1

This outlier correction is a function within a module, which is called by another function, because I have ~30 input files.

If I do this with the solution by @mozway then I want to save the mean of each column from each files, respectively, into a single file. But then the header (containing "Filename", "Time", "60Ni", "61Ni" and so on) is missing. What am I doing wrong?

This is the whole code from the outlier function:

import modules.config as conf
import os
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from scipy import stats
from pathlib import Path

outfile = (conf.outdir) 
infile = (conf.WorkspaceVariableInput)

def outlier_Cu_blk(file, append=True):

    global outfile   # get the path of the output folder
    
    """Create an output folder if it does not exist"""
    try:
        os.makedirs(outfile + '/blk')
    except FileExistsError:
        # directory already exists
        pass
    
    outfile_blk = outfile + '/blk' #open this folder
    fullname = os.path.join(outfile_blk, 'Cu_export_blk.csv') # Export file
    entries = Path(infile + '/Blk') 
    plot_name = Path(entries).stem
    basename = os.path.basename(file)
    
    """Reading of the input files"""
    
    pd.options.display.float_format = '{:.4f}'.format # Parameter for pandas

    data = pd.read_csv(file, sep='\t', names=['Time', '60Ni', '61Ni', '62Ni', '63Cu', '64Ni', '65Cu', '66Zn'], skiprows=6, nrows=30, index_col=False, dtype=float) # reading of the textfile, 6 rows are skipped and 30 rows read. Index column is deactivated
   
    """Outlier correction"""
    
    cols = list(data.drop(columns='Time').columns)
    datao = pd.DataFrame({'Time':data['Time']})
    datao[cols] = data[cols].where(np.abs(stats.zscore(data[cols])) < 2)

    """calculating the mean of the corrected data and save into a single file"""
    
    datao.to_csv('Cu_export_blk.csv', sep='\t', header = True, index_label='Index_name')
    
    mean_filtered_transposed = pd.DataFrame(data=np.mean(data)).T
    mean_filtered_transposed['Time'] = pd.to_datetime(mean_filtered_transposed["Time"], unit='s')
    clean = mean_filtered_transposed.drop(mean_filtered_transposed.columns[[0]], axis=1) 
    clean.insert(0, 'Inputfile', file)
    print(mean_filtered_transposed)
    print(clean)
    if append:
        clean.to_csv(fullname, sep=' ', mode="a", header=False, index_label='Index_name')
    else:
        clean.to_csv(fullname, sep=' ', mode="w", header=True, index_label='Index_name')

I know it may be quite a mess, but I am just a beginner :-)

Answer 2

If you need to replace outliers by missing values, use DataFrame.mask:

df = df.mask(np.abs(stats.zscore(df)) < 2)

#working for replace outlier by missing values
#df = df.mask(np.abs(stats.zscore(df)) < 2, np.nan)

I just want to get this single value deleted.

This is not possible, we can only remove row(s) like your solution.

Answer 3

It would be better to provide your data, but IIUC, use mask to mask your outliers with NaN:

from scipy import stats
cols = list(df.drop(columns='Time').columns)
# or
# cols = ['60Ni', '61Ni', '62Ni', '63Cu', '64Ni', '65Cu', '66Zn']

df[cols] = df[cols].mask(np.abs(stats.zscore(df[cols])) >= 2)

or with where

from scipy import stats
cols = list(df.drop(columns='Time').columns)
# or
# cols = ['60Ni', '61Ni', '62Ni', '63Cu', '64Ni', '65Cu', '66Zn']

df[cols] = df[cols].where(np.abs(stats.zscore(df[cols])) < 2)

output:

         Time      60Ni      61Ni      62Ni      63Cu      64Ni      65Cu      66Zn
0    0.000000  0.004247  0.000455  0.000836  0.003215  0.003216  0.001596  0.001984
1    1.052000  0.004264  0.000517  0.000829  0.003155  0.003217  0.001623  0.001874
2    2.103000  0.004275  0.000480  0.000832  0.003212  0.003147  0.001623  0.001880
3    3.155000  0.004279  0.000483  0.000797  0.003219  0.003230  0.001604  0.001938
4    4.207000  0.004212  0.000442  0.000801  0.003243  0.003167  0.001591  0.001904
5    5.258000  0.004268  0.000513  0.000831  0.003144  0.003117  0.001604  0.001815
6    6.310000  0.004183  0.000505  0.000790  0.003131  0.003101  0.001570  0.001818
7    7.361000  0.004296  0.000491  0.000891  0.003204  0.003028  0.001533  0.001788
8    8.413000  0.004336       NaN  0.000863       NaN  0.002987  0.001608  0.001796
9    9.464000  0.004290  0.000484  0.000853  0.003223  0.003006  0.001584  0.001700
10  10.516000  0.004288  0.000550       NaN  0.003219  0.002950       NaN  0.001783
11  11.566999  0.004260  0.000467  0.000774  0.003194  0.002854  0.001569  0.001737
12  12.618999  0.004265  0.000501  0.000861  0.003218  0.002866  0.001595  0.001659
13  13.671000  0.004223  0.000465  0.000863  0.003159  0.002802  0.001629  0.001673
14  14.722000  0.004234  0.000488  0.000832  0.003106  0.002766  0.001579  0.001671
15  15.774000  0.004264  0.000527  0.000834  0.003151  0.002748  0.001522  0.001639
16  16.825001  0.004173  0.000515  0.000785  0.003132  0.002736       NaN  0.001647
17  17.876999  0.004210  0.000458  0.000798  0.003183  0.002714  0.001605  0.001607
18  18.929001  0.004215  0.000492  0.000850  0.003178  0.002589  0.001561  0.001608
19  19.980000  0.004172  0.000444  0.000845  0.003142  0.002649  0.001588  0.001548
20  21.031000  0.004235  0.000509  0.000822  0.003190  0.002646  0.001557  0.001516
21  22.083000  0.004160  0.000521  0.000779  0.003094  0.002504  0.001598  0.001551
22  23.133999  0.004095  0.000528  0.000816  0.003165  0.002606  0.001514  0.001546
23  24.186001  0.004191  0.000474  0.000825  0.003078  0.002458  0.001617  0.001503
24  25.237999  0.004156  0.000448  0.000801  0.003119  0.002549  0.001551  0.001538
25  26.289000       NaN  0.000427  0.000825       NaN  0.002365  0.001566  0.001419
26  27.341000  0.004161  0.000464  0.000841  0.003150  0.002621  0.001559  0.001535
27  28.392000  0.004124  0.000542  0.000831  0.003129  0.002427  0.001607  0.001476
28  29.444000  0.004186  0.000499       NaN  0.003080  0.002371  0.001568  0.001444
29  30.496000  0.004092  0.000532  0.000837  0.003113  0.002385  0.001580  0.001434