CODEWITHSUNDEEP
pythonpandasrolling-computation
Thedudeabides
Thedudeabides

Reputation: 93

Rolling stats pandas with based of fixed period or available data

I would like to take rolling stats over a fixed period say 5 day, eg 

                 DATE       Price
ID          
AAPL US Equity  2015-01-02  109.33
AAPL US Equity  2015-01-05  106.25
AAPL US Equity  2015-01-06  106.26
AAPL US Equity  2015-01-07  107.75
AAPL US Equity  2015-01-08  111.89
AAPL US Equity  2015-01-09  112.01
AAPL US Equity  2015-01-12  109.25
AAPL US Equity  2015-01-13  110.22
AAPL US Equity  2015-01-14  109.80
AAPL US Equity  2015-01-15  106.82

doing


df['Average']=df['Price'].rolling(5).mean()

to give 

                DATE        Price   Average
ID                  
AAPL US Equity  2015-01-02  109.33  NaN
AAPL US Equity  2015-01-05  106.25  NaN
AAPL US Equity  2015-01-06  106.26  NaN
AAPL US Equity  2015-01-07  107.75  NaN
AAPL US Equity  2015-01-08  111.89  108.296
AAPL US Equity  2015-01-09  112.01  108.832
AAPL US Equity  2015-01-12  109.25  109.432
AAPL US Equity  2015-01-13  110.22  110.224
AAPL US Equity  2015-01-14  109.80  110.634
AAPL US Equity  2015-01-15  106.82  109.620

how can i modify to apply any function to get any rolling stats over the fixed period but for the first few rows over the available data ie for first row average will be first day price, for first two rows average will be first two day price etc

I know I can use iterrows to achieve this in this one of 'average' value case, but ideally I would like to use it for any stat like quantile,std etc

Something like

df['Average']=my_rolling_stat(df['Price'],period=5,function='mean')
df['Stdev']=my_rolling_stat(df['Price'],period=10,function='std')
df['95_Perc']=my_rolling_stat(df['Price'],period=10,function='quantile',quantile_value=0.95)

Thanks in advance

Upvotes: 1

Views: 94

Answers (2)

Code Different
Code Different

Reputation: 93141

To add to Scott Boston's answer, you can define your rolling stat function as:

def my_rolling_stat(series, period, function, **kwargs):
    window = series.rolling(period, min_periods=1)
    func = getattr(window, function)

    return func(**kwargs)

Usage:

my_rolling_stat(df['Price'], period=5, function='mean')
my_rolling_stat(df['Price'], period=10, function='std')
my_rolling_stat(df['Price'], period=10, function='quantile', quantile=0.95)

The list of functions and their arguments can be found from the Computation / Descriptive stats list.

Upvotes: 2

Scott Boston
Scott Boston

Reputation: 153460

IIUC, use min_periods parameter in rolling:

df['Average']=df['Price'].rolling(5, min_periods=1).mean()

Output:

0    109.3300
1    107.7900
2    107.2800
3    107.3975
4    108.2960
5    108.8320
6    109.4320
7    110.2240
8    110.6340
9    109.6200
Name: PRICE, dtype: float64

Upvotes: 4

Related Questions