I want to perform a moving window linear fit to the columns in my dataframe.
n =5
df = pd.DataFrame(index=pd.date_range('1/1/2000', periods=n))
df['B'] = [1.9,2.3,4.4,5.6,7.3]
df['A'] = [3.2,1.3,5.6,9.4,10.4]
B A
2000-01-01 1.9 3.2
2000-01-02 2.3 1.3
2000-01-03 4.4 5.6
2000-01-04 5.6 9.4
2000-01-05 7.3 10.4
For, say, column B, I want to perform a linear fit using the first two rows, then another linear fit using the second and third rown and so on. And the same for column A. I am only interested in the slope of the fit so at the end, I want a new dataframe with the entries above replaced by the different rolling slopes.
After doing
df.reset_index()
I try something like
model = pd.ols(y=df['A'], x=df['index'], window_type='rolling',window=3)
But I get
KeyError: 'index'
EDIT:
I aded a new column
df['i'] = range(0,len(df))
and I can now run
pd.ols(y=df['A'], x=df.i, window_type='rolling',window=3)
(it gives an error for window=2)
I am not understaing this well because I was expecting a string of numbers but I get just one result
-------------------------Summary of Regression Analysis---------------
Formula: Y ~ <x> + <intercept>
Number of Observations: 3
Number of Degrees of Freedom: 2
R-squared: 0.8981
Adj R-squared: 0.7963
Rmse: 1.1431
F-stat (1, 1): 8.8163, p-value: 0.2068
Degrees of Freedom: model 1, resid 1
-----------------------Summary of Estimated Coefficients--------------
Variable Coef Std Err t-stat p-value CI 2.5% CI 97.5%
--------------------------------------------------------------------------------
x 2.4000 0.8083 2.97 0.2068 0.8158 3.9842
intercept 1.2667 2.5131 0.50 0.7028 -3.6590 6.1923
---------------------------------End of Summary---------------------------------
EDIT 2:
Now I understand better what is going on. I can acces the different values of the fits using
model.beta
I havent tried it out, but I don't think you need to specify the window_type='rolling', if you specify the window to something, window will automatically be set to rolling.
Source.
I have problems doing this with the DatetimeIndex you created with pd.date_range, and find datetimes a confusing pain to work with in general due to the number of types out there and apparent incompatibility between APIs. Here's how I would do it if the date were an integer (e.g. days since 12/31/99, or years) or float in your example. It won't help your datetime problem, but hopefully it helps with the rolling linear fit part.
Generating your date with integers instead of datetimes:
df = pd.DataFrame()
df['date'] = range(1,6)
df['B'] = [1.9,2.3,4.4,5.6,7.3]
df['A'] = [3.2,1.3,5.6,9.4,10.4]
date B A
0 1 1.9 3.2
1 2 2.3 1.3
2 3 4.4 5.6
3 4 5.6 9.4
4 5 7.3 10.4
Since you want to group by 2 dates every time, then fit a linear model on each group, let's duplicate the records and number each group with the index:
df_dbl = pd.concat([df,df], names = ['date', 'B', 'A']).sort()
df_dbl = df_dbl.iloc[1:-1] # removes the first and last row
date B A
0 1 1.9 3.2 # this record is removed
0 1 1.9 3.2
1 2 2.3 1.3
1 2 2.3 1.3
2 3 4.4 5.6
2 3 4.4 5.6
3 4 5.6 9.4
3 4 5.6 9.4
4 5 7.3 10.4
4 5 7.3 10.4 # this record is removed
c = df_dbl.index[1:len(df_dbl.index)].tolist()
c.append(max(df_dbl.index))
df_dbl.index = c
date B A
1 1 1.9 3.2
1 2 2.3 1.3
2 2 2.3 1.3
2 3 4.4 5.6
3 3 4.4 5.6
3 4 5.6 9.4
4 4 5.6 9.4
4 5 7.3 10.4
Now it's ready to group by index to run linear models on B vs. date, which I learned from Using Pandas groupby to calculate many slopes. I use scipy.stats.linregress since I got weird results with pd.ols and couldn't find good documentation to understand why (perhaps because it's geared toward datetime).
1 0.4
2 2.1
3 1.2
4 1.7
Related
I have a data frame with columns containing different country values, I would like to have a function that shifts the rows in this dataframe independently without the dates. For example, I have a list of related profile shifters for each country which would be used in shifting the rows.
If the profile shifter for a country is -3, that country column, is shifted 3 times downwards, while the last 3 values become the first 3 values in the dataframe. If a profile shifter is +3, the third value of a row is shifted upwards while the first 2 values become the last values in that column.
After the rows have been shifted instead of having the default Nan value appear in the empty cells, I want the preceding or succeeding values to take up the empty cells. The function should also return a data frame Sample-dataset Profile Shifter Expected-results.
Sample Dataset:
Datetime ARG AUS BRA
1/1/2050 0.00 0.1 2.1 3.1
1/1/2050 1.00 0.2 2.2 3.2
1/1/2050 2.00 0.3 2.3 3.3
1/1/2050 3.00 0.4 2.4 3.4
1/1/2050 4.00 0.5 2.5 3.5
1/1/2050 5.00 0.6 2.6 3.6
Country Profile Shifters:
Country ARG AUS BRA
UTC -3 -2 4
Desired Output:
Datetime ARG AUS BRA
1/1/2050 0.00 0.3 2.4 3.4
1/1/2050 1.00 0.4 2.5 3.5
1/1/2050 2.00 0.5 2.1 3.1
1/1/2050 3.00 0.1 2.2 3.2
1/1/2050 4.00 0.2 2.3 3.3
This is what I have been trying for days now but it's not working
cols = df1.columns
for i in cols:
if i == 'ARG':
x = df1.iat[0:3,0]
df1['ARG'] = df1.ARG.shift(periods=-3)
df1['ARG'].replace(to_replace=np.nan, x)
elif i == 'AUS':
df1['AUS'] = df1.AUS.shift(periods=2)
elif i == 'BRA':
df1['BRA'] = df1.BRA.shift(periods=1)
else:
pass
This works but is far from being 'good pandas'. I hope that someone will come along and give a nicer, cleaner 'more pandas' answer.
Imports used:
import pandas as pd
import datetime as datetime
Offset data setup:
offsets = pd.DataFrame({"Country" : ["ARG", "AUS", "BRA"], "UTC Offset" : [-3, -2, 4]})
Produces:
Country UTC Offset
0 ARG -3
1 AUS -2
2 BRA 4
Note that the timezone offset data I've used here is in a slightly different structure from the example data (country codes by rows, rather than columns). Also worth pointing out that Australia and Brazil have several time zones, so there is no one single UTC offset which applies to those whole countries (only one in Argentina though).
Sample data setup:
sampleDf = pd.DataFrame()
for i in range(6):
dt = datetime.datetime(2050,1,1,i)
sampleDf = sampleDf.append({'Datetime' : dt,
'ARG' : i / 10,
'AUS' : (i + 10)/ 10,
'BRA' : (i + 20) / 10},
ignore_index=True)
Produces:
Datetime ARG AUS BRA
0 2050-01-01 00:00:00 0.0 1.0 2.0
1 2050-01-01 01:00:00 0.1 1.1 2.1
2 2050-01-01 02:00:00 0.2 1.2 2.2
3 2050-01-01 03:00:00 0.3 1.3 2.3
4 2050-01-01 04:00:00 0.4 1.4 2.4
5 2050-01-01 05:00:00 0.5 1.5 2.5
Code to shift cells:
for idx, offsetData in offsets.iterrows(): # See note 1
countryCode = offsetData["Country"]
utcOffset = offsetData["UTC Offset"]
dfRowCount = sampleDf.shape[0]
wrappedOffset = (dfRowCount + utcOffset) if utcOffset < 0 else \
(-dfRowCount + utcOffset) # See note 2
countryData = sampleDf[countryCode]
sampleDf[countryCode] = pd.concat([countryData.shift(utcOffset).dropna(),
countryData.shift(wrappedOffset).dropna()]).sort_index() # See note 3
Produces:
Datetime ARG AUS BRA
0 2050-01-01 00:00:00 0.0 1.4 2.4
1 2050-01-01 01:00:00 0.1 1.5 2.5
2 2050-01-01 02:00:00 0.2 1.0 2.0
3 2050-01-01 03:00:00 0.3 1.1 2.1
4 2050-01-01 04:00:00 0.4 1.2 2.2
5 2050-01-01 05:00:00 0.5 1.3 2.3
Notes
Iterating over rows in pandas like this (to me) indicates 'you've run out of pandas skill, and are kind of going against the design of pandas'. What I have here works, but it won't benefit from any/many of the efficiencies of using pandas, and would not be appropriate for a large dataset. Using itertuples rather than iterrows is supposed to be quicker, but I think neither is great, so I went with what seemed most readable for this case.
This solution does two shifts, one of the data shifted by the timezone offset, then a second shift of everything else to fill in what would otherwise be NaN holes left by the first shift. This line calculates the size of that second shift.
Finally, the results of the two shifts are concatenated together (after dropping any NaN values from both of them) and assigned back to the original (unshifted) column. sort_index puts them back in order based on the index, rather than having the two shifted parts one-after-another.
I have a dataframe of portfolio returns:
date Portfolio %
30/11/2001 4.8
31/12/2001 -0.7
31/01/2002 1.3
28/02/2002 -1.4
29/03/2002 3.3
I need to create an index of returns, but to do this i need to have a starting figure of 1.0 and the formula references the previous row. The output should look like this:
date Portfolio % Index
1.0 NaN
30/11/2001 4.8 1.048
31/12/2001 -0.7 1.040
31/01/2002 1.3 1.054
28/02/2002 -1.4 1.039
29/03/2002 3.3 1.073
As an example the formula for the second result is:
1.048*(1+-0.7/100)
I've tried the following code, but it doesn't get the required result.
portfolio['Index'] = portfolio['Portfolio %'] / portfolio['Portfolio %'].iloc[0]
The issues i have:
I can't get the starting variable
I can't get the formula to reference the previous row.
I believe it is the same issue as this post: Create and index from returns PANDAS. However, it was never answered fully.
Use, Series.div, Series.add along with Series.cumprod :
df['Index'] = df['Portfolio %'].div(100).add(1).cumprod()
Result:
# print(df)
date Portfolio % Index
0 30/11/2001 4.8 1.048000
1 31/12/2001 -0.7 1.040664
2 31/01/2002 1.3 1.054193
3 28/02/2002 -1.4 1.039434
4 29/03/2002 3.3 1.073735
This question already has answers here:
Reshape wide to long in pandas
(2 answers)
Closed 4 years ago.
I am manipulating a data frame using Pandas in Python to match a specific format.
I currently have a data frame with a row for each measurement location (A or B). Each row has a nominal target and multiple measured data points.
This is the format I currently have:
df=
Location Nominal Meas1 Meas2 Meas3
A 4.0 3.8 4.1 4.3
B 9.0 8.7 8.9 9.1
I need to manipulate this data so there is only one measured data point per row, and copy the Location and Nominal values from the source rows to the new rows. The measured data also needs to be put in the first column.
This is the format I need:
df =
Meas Location Nominal
3.8 A 4.0
4.1 A 4.0
4.3 A 4.0
8.7 B 9.0
8.9 B 9.0
9.1 B 9.0
I have tried concat and append functions with and without transpose() with no success.
This is the most similar example I was able to find, but it did not get me there:
for index, row in df.iterrows():
pd.concat([row]*3, ignore_index=True)
Thank you!
Its' a wide to long problem
pd.wide_to_long(df,'Meas',i=['Location','Nominal'],j='drop').reset_index().drop('drop',1)
Out[637]:
Location Nominal Meas
0 A 4.0 3.8
1 A 4.0 4.1
2 A 4.0 4.3
3 B 9.0 8.7
4 B 9.0 8.9
5 B 9.0 9.1
Another solution, using melt:
new_df = (df.melt(['Location','Nominal'],
['Meas1', 'Meas2', 'Meas3'],
value_name = 'Meas')
.drop('variable', axis=1)
.sort_values('Location'))
>>> new_df
Location Nominal Meas
0 A 4.0 3.8
2 A 4.0 4.1
4 A 4.0 4.3
1 B 9.0 8.7
3 B 9.0 8.9
5 B 9.0 9.1
I want to use the value of df.d to define the row from which I calculate the relative value by using the formula df.a/df.a[x] while x is defined as df.d. But somehow this doesn't work. My approach so far is this one:
import pandas as pd
import numpy as np
import datetime
randn = np.random.randn
rng = pd.date_range('1/1/2011', periods=10, freq='D')
df = pd.DataFrame({'a': [1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0], 'b': [1.1, 1.7, 1.3, 1.6, 1.5, 1.1, 1.5, 1.7, 2.1, 1.9],'c':[None] * 10},index=rng)
df["d"]= [0,0,0,0,4,4,4,4,8,8]
df["c"] =df.a/df.a[df.d]
All I get is the error: ValueError: cannot reindex from a duplicate axis
To clarify this: I want to set df.a/df.a[0] for the first 4 rows, df.a/df.a[4] for the next 4 and df.a/df.a[8] for the last 2 rows according to df["d"]= [0,0,0,0,4,4,4,4,8,8]
So how can I refer to a value in the dataframe correcty, without getting this error?
The output I seek looks like this:
a b c d
2011-01-01 1.1 1.1 1 0 # df.a/df.a[0]
2011-01-02 1.2 1.7 1.090909090909091 0 # df.a/df.a[0]
2011-01-03 1.3 1.3 1.181818181818182 0 # df.a/df.a[0]
2011-01-04 1.4 1.6 1.272727272727273 0 # df.a/df.a[0]
2011-01-05 1.5 1.5 1 4 # df.a/df.a[4]
2011-01-06 1.6 1.1 1.066666666666667 4 # df.a/df.a[4]
2011-01-07 1.7 1.5 1.133333333333333 4 # df.a/df.a[4]
2011-01-08 1.8 1.7 1.2 4 # df.a/df.a[4]
2011-01-09 1.9 2.1 1 8 # df.a/df.a[8]
2011-01-10 2.0 1.9 1.052631578947368 8 # df.a/df.a[8]
The pandas version used is 0.16.0
Thanks a lot for your support!
With regards to your original Error, I get a different error -
Unsupported Iterator Index. That's 'cos I am trying to get values from df.a at an index which is a series (df.d) and not an Index value. (I've pandas version 0.13.1), but to solve your actual problem -
Here's how I could go about it.
df['d'] = pd.Series([0,0,0,0,4,4,4,4,8,8], index=rng)
x = df.a.iloc[df.d]
note here - x you get has a different date index so simply
df['c'] = df.a/x # incorrect
won't work. we are only interested in values - so we take them out and assign (ignoring the index).
df['c'] = df.a/x.values # We ignore the index of 'x'
or as a short form
df['c'] = df.a/df.a.iloc[df.d].values
What is not clear to me yet is - even though the index of df.d is correct why simple df.a.iloc won't work.
Hope that helps.
You might want to use this instead of your last line:
df["c"] = df.a.values / df.a[df.d].values
print df
Which yields:
a b c d
2011-01-01 1.1 1.1 1.000 0
2011-01-02 1.2 1.7 1.091 0
2011-01-03 1.3 1.3 1.182 0
2011-01-04 1.4 1.6 1.273 0
2011-01-05 1.5 1.5 1.000 4
2011-01-06 1.6 1.1 1.067 4
2011-01-07 1.7 1.5 1.133 4
2011-01-08 1.8 1.7 1.200 4
2011-01-09 1.9 2.1 1.000 8
2011-01-10 2.0 1.9 1.053 8
The reason you had an error is because two series you tried to divide one by another had different indexes (not aligned ones). Adding .values gets rid of the indexes and solves the issue.
I want to apply a function f to many slices within each row of a pandas DataFrame.
For example, DataFrame df would look as such:
df = pandas.DataFrame(np.round(np.random.normal(size=(2,49)), 2))
So, I have a dataframe of 2 rows by 49 columns, and my function needs to be applied to every consequent slice of 7 data points in both rows, and so that the resulting dataframe looks identical to the input dataframe.
I was doing it as such:
df1=df.copy()
df1.T[:7], df1.T[7:14], df1.T[14:21],..., df1.T[43:50] = f(df.T.iloc[:7,:]), f(df.T.iloc[7:14,:]),..., f(df.T.iloc[43:50,:])
As you can see that's a whole lot of redundant code.. so I would like to create a loop or something so that it applies the function to every 7 subsequent data point...
I have no idea how to approach this. Is there a more elegant way to do this?
I thought I could maybe use a transform function for this, but in the pandas documentation I can only see that applied to a dataframe that has been grouped and not on slices of the data....
Hopefully this is clear.. let me know.
Thank you.
To avoid redundant code you can just do a loop like this:
STEP = 7
for i in range(0,len(df),STEP):
df1.T[i:i+STEP] = f(df1.T[i:i+STEP]) # could also do an apply here somehow, depending on what you want to do
Don't Repeat Yourself
You don't provide any examples of your desired output, so here's my best guess at what you want...
If your data are lumped into groups of seven, the you need to come up with a way to label them as such.
If other words, you with want to work with arbitrary arrays, use numpy. If you want to work with labeled, meaningful data and it's associated metadata, then use pandas.
Also, pandas works more efficiently when operating (and displaying!) row-wise data. So that mean store data long (49x2), not wide (2x49)
Here's an example of what I mean. I have the same 49x2 random array, but assigned grouping labels to the rows ahead of time.
Let's yeah you're reading in some wide-ish data as following:
import pandas
import numpy
from io import StringIO # python 3
# from StringIO import StringIO # python 2
datafile = StringIO("""\
A,B,C,D,E,F,G,H,I,J
0.0,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9
1.0,1.1,1.2,1.3,1.4,1.5,1.6,1.7,1.8,1.9
2.0,2.1,2.2,2.3,2.4,2.5,2.6,2.7,2.8,2.9
""")
df = pandas.read_csv(datafile)
print(df)
A B C D E F G H I J
0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
1 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
2 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9
You could add a cluster value to the columns, like so:
cluster_size = 3
col_vals = []
for n, col in enumerate(df.columns):
cluster = int(n/cluster_size)
col_vals.append((cluster, col))
df.columns = pandas.Index(col_vals)
print(df)
0 1 2 3
A B C D E F G H I J
0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
1 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
2 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9
By default, the groupby method tries to group rows, but you can group columns (I just fogured this out), by passing axis=1 when you create the object. So the sum of each cluster of columns for each row is as follows:
df.groupby(axis=1, level=0).sum()
0 1 2 3
0 0.3 1.2 2.1 0.9
1 3.3 4.2 5.1 1.9
2 6.3 7.2 8.1 2.9
But again, if all you're doing is more "global" operations, there's no need to any of this.
In-place column cluster operation
df[0] *= 5
print(df)
0 1 2 3
A B C D E F G H I J
0 0 2.5 5 0.3 0.4 0.5 0.6 0.7 0.8 0.9
1 25 27.5 30 1.3 1.4 1.5 1.6 1.7 1.8 1.9
2 50 52.5 55 2.3 2.4 2.5 2.6 2.7 2.8 2.9
In-place row operation
df.T[0] += 20
0 1 2 3
A B C D E F G H I J
0 20 22.5 25 20.3 20.4 20.5 20.6 20.7 20.8 20.9
1 25 27.5 30 1.3 1.4 1.5 1.6 1.7 1.8 1.9
2 50 52.5 55 2.3 2.4 2.5 2.6 2.7 2.8 2.9
Operate on the entire dataframe at once
def myFunc(x):
return 5 + x**2
myFunc(df)
0 1 2 3
A B C D E F G H I J
0 405 511.25 630 417.09 421.16 425.25 429.36 433.49 437.64 441.81
1 630 761.25 905 6.69 6.96 7.25 7.56 7.89 8.24 8.61
2 2505 2761.25 3030 10.29 10.76 11.25 11.76 12.29 12.84 13.41