# How to calculate rows where it has values in at least 3 columns

• Python

## Main Question or Discussion Point

I am trying to count number of rows that has values in at least 3 columns so the output based on the image shared should be 4. I tried using the code below but it is resulting in the same shape as the whole table which is 7.

Python:
counts = df[(df[['a', 'b', 'c', 'd]] != 'no_label').count(axis=1) >= 3]

#### Attachments

• 2.3 KB Views: 296

Related Programming and Computer Science News on Phys.org
mfb
Mentor
sum(row.count("no_label")<2 for row in df)
If your table includes the header but you don't want to count that: subtract 1.

thank you for this suggestion. but what if i only want to check from these columns a,b,c, and d? as I have other columns e,f,g that also has the 'no_label' value but I don't want to consider them,

mfb
Mentor
sum(row[0:4].count("no_label")<2 for row in df)

i wanted to do it this way so that i can select the columns that could be in other position in the dataframe:

Python:
cols = ['a', 'b', 'c', 'd'']

sum(row[cols].count("no_label") < 2 for row in df)
but it gives me this error :

TypeError: string indices must be integers, not list

what can i do to use the column names explicitly instead. thanks

mfb
Mentor
Where do the column labels come from? Are they the first row of you array? I wouldn't do that, it is bad style. Select them based on integers.
If you absolutely need strings convert them to integers for the sum.

yes they are but in this case there are more than 10 columns which makes it difficult to determine the location of the column using integers.

mfb
Mentor
You keep changing the task. Why don't you show what exactly you want to do?

cols={'a':0,'b':1,'c':2,'d':3}
sum(row[cols['a']:cols['c']].count("no")>0 for row in df)

Or, if the first row of df has the column names:
Code:
cols2={}
for x,y in enumerate(df):
cols2[y]=x

• msn009 and Tom.G
thanks. sorry for not explaining the scenario in detail.