I've got a 3-column dataset with 7100 rows.

data.isna().sum() shows, that one column contains 117 NaN vlaues, the others 0. data.isnull().sum() shows also 117 for one and 0 for the other columns.

data.dropna(inplace=True) drops 351 rows. Can anyone explain this to me? Am I doing anything wrong?

Edit:

I now examined the deleted rows. There are 351 rows deleted, where dropped.isna().sum().sum() shows a total of 117 NaN values.

dropped[~dropped['description'].isna()] shows an empty table. So the result seems to be correct as far as I can see.

Now I'm just curious how the difference in counting occurs.

Sadly I'm not able/allowed to provide a data sample.

CodePudding user response：

data.isna().sum() returns the total number of NAN values in your dataframe and using data.dropna() will drop all NAN values. You can specifically check the number of NAN values by creating a subset for example: nan_rows=dataframe[dataframe.columnNameWithNanValues.isna()] to check for the NAN values and then return the shape of your dataframe. Next, use .dropna() without the inplace=True argument to drop NAN and Null values.

CodePudding user response：

Found the solution. pretty simple... I've got three columns, one column contains 117 NaN values. 117 values for 3 columns are a total of 351 fields to be deleted. Since i used the df.size to measure the deleted size, which counts fields and not rows, I got 351 "deleted fields", which is totally correct.