CODEWITHSUNDEEP
pandas
alernerdev
alernerdev

Reputation: 2064

How to determine column values changing in a particular direction?

I have a dataframe consisting of 1 column of periods (year and quarters) and another column of some productivity numbers for that period. My task is to identify a period where, for example, I have two consecutive quarters of productivity decline; or, similarly, two consecutive quarters of growth. I imagine I can use brute force and just loop over the rows looking at several rows at a time, but I was reading that this might have something to do with a "shift" function -- but I dont understand how it works. Thank you for your help

    1971q1  1,137.8
    1971q2  1,159.4
    1971q3  1,180.3
    1971q4  1,173.6
    1972q1  1,163.8
    1972q2  1,140.1
    1972q3  1,145.8
    1972q4  1,150.0

Upvotes: 1

Views: 547

Answers (2)

MaxU - stand with Ukraine
MaxU - stand with Ukraine

Reputation: 210832

It's still not quite clear to me how your desired data set should look like.

Here is a method which helps to identify grows and drops:

In [450]: df
Out[450]:
   period     val
0  1971q1  1137.8
1  1971q2  1159.4
2  1971q3  1180.3
3  1971q4  1173.6
4  1972q1  1163.8
5  1972q2  1140.1
6  1972q3  1145.8
7  1972q4  1150.0

In [451]: np.sign(df.val.diff().fillna(0))
Out[451]:
0    0.0
1    1.0
2    1.0
3   -1.0
4   -1.0
5   -1.0
6    1.0
7    1.0
Name: val, dtype: float64

In [452]: df.loc[np.sign(df.val.diff().fillna(0)) < 0]
Out[452]:
   period     val
3  1971q4  1173.6
4  1972q1  1163.8
5  1972q2  1140.1

Upvotes: 1

ℕʘʘḆḽḘ
ℕʘʘḆḽḘ

Reputation: 19375

try this buddy

#define a growth rate
df['growth_rate'] = np.log(df.production) - np.log(df.production).shift(1)
#a recession is when there have been two quarters of negative growth. 
df['recession'] =  (df['growth_rate'] < 0 ) &  (df['growth_rate'].shift(1) < 0 )

Upvotes: 2

Related Questions