Use shift.
df['dA'] = df['A'] - df['A'].shift(-1)
Question
Lets say I have a dataframe like this
A B
0 a b
1 c d
2 e f
3 g h
0,1,2,3 are times, a, c, e, g is one time series and b, d, f, h is another time series. I need to be able to add two columns to the orignal dataframe which is got by computing the differences of consecutive rows for certain columns.
So i need something like this
A B dA
0 a b (a-c)
1 c d (c-e)
2 e f (e-g)
3 g h Nan
I saw something called diff on the dataframe/series but that does it slightly differently as in first element will become Nan.
Solution
Use shift.
df['dA'] = df['A'] - df['A'].shift(-1)
OTHER TIPS
You could use diff
and pass -1
as the periods
argument:
>>> df = pd.DataFrame({"A": [9, 4, 2, 1], "B": [12, 7, 5, 4]})
>>> df["dA"] = df["A"].diff(-1)
>>> df
A B dA
0 9 12 5
1 4 7 2
2 2 5 1
3 1 4 NaN
[4 rows x 3 columns]
When using data in CSV, this would work perfectly:
my_data = pd.read_csv('sale_data.csv')
df = pd.DataFrame(my_data)
df['New_column'] = df['target_column'].diff(1)
print(df) #for the console but not necessary
Rolling differences can also be calculated this way:
df=pd.DataFrame(my_data)
my_data = pd.read_csv('sales_data.csv')
i=0
j=1
while j < len(df['Target_column']):
j=df['Target_column'][i+1] - df['Target_column'][i] #the difference btwn two values in a column.
i+=1 #move to the next value in the column.
j+=1 #next value in the new column.
print(j)