CODEWITHSUNDEEP
pythonpandasnormalization
Niccola Tartaglia
Niccola Tartaglia

Reputation: 1667

Reverse z Score pandas dataframe

I am using this to compute the z score of my dataframe:

df_z=df.apply(zscore)

Is there a reverse operation that can give me the orginal values?

Upvotes: 6

Views: 3633

Answers (2)

Mike
Mike

Reputation: 1

You can use StandardScaler from SciKit Learn which basically will do the above logic for you. As the name does not indicate StandardScaler is Z Scoring class.

so

from sklearn.preprocessing import StandardScaler

zscore = StandardScaler()

# Do the scoring
out = zscore.fit_transform(x)

# Do the reverse
x = zscore.inverse_transform(out)

Upvotes: 0

kalidurge
kalidurge

Reputation: 289

There is no built-in way to go from df_z (z scores) back to df (original values). However, you can do it fairly easily as follows:

Step 1: Keep track of the mean and standard deviations of all of the original variables. Perhaps like this:

mean_std={}
for var in df.columns:
    mean_std[var]=(df[var].mean(), df[var].std())

Step 2: Convert back to z scores

def reverse_zscore(pandas_series, mean, std):
    '''Mean and standard deviation should be of original variable before standardization'''
    yis=pandas_series*std+mean
    return yis

original_mean, original_std = mean_std[var]
original_var_series = reverse_zscore(df_z[var], original_mean, original_std)

Alternatively, just store your original dataframe somewhere

Upvotes: 4

Related Questions