CODEWITHSUNDEEP
pandasindexingrowsdrop
user2100039
user2100039

Reputation: 1356

How to Drop Pandas Rows Based on Index Condition (True/False)

I need to apply simply an index constructed from the condition of "NaN" from one dataframe "df1" to a different dataframe "df2" so that all the rows as indicated by "True" in the index are dropped from df1 and df2 dataframes.

the series rowswithnan index looks like this:

0   False
1   True
2   True
3   False

and df1 looks like this:

   plant_name  power_kwh  hour  day  month  year
0  AREC        32963.4    23   31     12    2020
1  AREC        35328.2    22   31     12    2020
2  AREC        37523.6    21   31     12    2020
3  AREC        36446.0    20   31     12    2020

and df2 looks like this:

   plant_name  power_obs_kwh  hour  day  month  year
0  AREC        31548.4        23   31     12    2020
1  AREC        34732.2        22   31     12    2020
2  AREC        39023.7        21   31     12    2020
3  AREC        34946.0        20   31     12    2020

Based on the "True" conditions of the index "rowswithnan", i need to delete or drop rows for both df1 and df2 such that they look like this for df2:

   plant_name  power_obs_kwh  hour  day  month  year
0  AREC        31548.4        23   31     12    2020
3  AREC        34946.0        20   31     12    2020

and df1, the same like this:

   plant_name  power_kwh  hour  day  month  year
0  AREC        32963.4    23   31     12    2020
3  AREC        36446.0    20   31     12    2020

I have tried something like this with an error:

df1.drop(rowswithnan,inplace=True,axis=1)

KeyError: '[False False False ... True True True] not found in axis'

thank you!

Upvotes: 0

Views: 766

Answers (2)

tdy
tdy

Reputation: 41327

Boolean indexing is simpler here:

df1 = df1[~rowswithnan]
df2 = df2[~rowswithnan]

It's also possible to use your original drop() method, but drop() wants the labels so we end up using boolean indexing anyway to get the index:

df1.drop(df1[rowswithnan].index, inplace=True)
df2.drop(df2[rowswithnan].index, inplace=True)

Upvotes: 1

Ananay Mital
Ananay Mital

Reputation: 1475

This is what you need. You need to make sure that number of rows of df1, df2 and rowswithnan are same.

df1[~rowswithnan]
df2[~rowswithnan]

More about Boolean indexing

Upvotes: 1

Related Questions