Pandas pct change from initial value

Question

I want to find the pct_change of Dew_P Temp (C) from the initial value of -3.9. I want the pct_change in a new column.

Source here:

weather = pd.read_csv('https://raw.githubusercontent.com/jvns/pandas-cookbook/master/data/weather_2012.csv')
weather[weather.columns[:4]].head()


Date/Time        Temp (C)   Dew_P Temp (C)  Rel Hum (%)
0   2012-01-01   -1.8       -3.9             86
1   2012-01-01   -1.8       -3.7             87
2   2012-01-01   -1.8       -3.4             89
3   2012-01-01   -1.5       -3.2             88
4   2012-01-01   -1.5       -3.3             88

I have tried variations of this for loop (even going as far as adding an index shown here) but to no avail:

for index, dew_point in weather['Dew_P Temp (C)'].iteritems():
    new = weather['Dew_P Temp (C)'][index]
    old = weather['Dew_P Temp (C)'][0]
    pct_diff = (new-old)/old*100
    weather['pct_diff'] = pct_diff

I think the problem is the weather['pct_diff'] , it doesn't take the new it takes the last value of the data frame and subtracts it from old

So its always (2.1-3.9)/3.9*100 thus my percent change is always -46%.

The end result I want is this:

Date/Time        Temp (C)   Dew_P Temp (C)  Rel Hum (%)   pct_diff
0   2012-01-01   -1.8       -3.9             86           0.00%
1   2012-01-01   -1.8       -3.7             87           5.12%
2   2012-01-01   -1.8       -3.4             89           12.82%  

Any ideas? Thanks!

Answer 1

You can use iat to access the scalar value (eg iat[0] accesses the first value in the series).

df = weather
df['pct_diff'] = df['Dew_P Temp (C)'] / df['Dew_P Temp (C)'].iat[0] - 1

Answer 2

IIUC you can do it this way:

In [88]: ((weather['Dew Point Temp (C)'] - weather.ix[0, 'Dew Point Temp (C)']).abs() / weather.ix[0, 'Dew Point Temp (C)']).abs() * 100
Out[88]:
0         0.000000
1         5.128205
2        12.820513
3        17.948718
4        15.384615
5        15.384615
6        20.512821
7         7.692308
8         7.692308
9        20.512821

Answer 3

I find this more graceful

weather['Dew_P Temp (C)'].pct_change().fillna(0).add(1).cumprod().sub(1)

0    0.000000
1   -0.051282
2   -0.128205
3   -0.179487
4   -0.153846
Name: Dew_P Temp (C), dtype: float64

---

To get your expected output with absolute values

weather['pct_diff'] = weather['Dew_P Temp (C)'].pct_change().fillna(0).add(1).cumprod().sub(1).abs()
weather