Python for Finance: Analyze Big Financial Data

                                    No2             0.049462

                                    No3             0.570157

                                    No4         -0.327594

                                    dtype:  float64

In  [ 29 ]: df.cumsum()

Out[29]:                                                                        No1                         No2                         No3                         No4

                                    2015-01-31  -0.737304       1.065173        0.073406        1.301174

                                    2015-02-28  -1.526122       0.079354        0.477201    -0.452609

                                    2015-03-31  -1.682003   -1.673318       1.514645    -0.853403

                                    2015-04-30  -2.459549       0.056960        1.931759    -0.669323

                                    2015-05-31  -4.223209   -0.318508       2.030438    -2.223147

                                    2015-06-30  -5.357467       1.083313        3.257562    -1.243758

                                    2015-07-31  -4.898629       0.940126        4.823263    -3.329621

                                    2015-08-31  -5.001687       0.573956        4.345227    -3.362430

                                    2015-09-30  -3.961370       0.445156        5.131414    -2.948346

There is also a shortcut to a number of often-used statistics for numerical data sets, the

describe method:

In  [ 30 ]: df.describe()

Out[30]:                                                    No1                         No2                         No3                         No4

                                    count       9.000000        9.000000        9.000000        9.000000

                                    mean        -0.440152       0.049462        0.570157    -0.327594

                                    std             0.847907        1.141676        0.642904        1.219345

                                    min         -1.763660   -1.752672   -0.478036   -2.085863

                                    25%         -0.788818   -0.375469       0.098678    -1.553824

                                    50%         -0.737304   -0.143187       0.417114    -0.032810

                                    75%         -0.103058       1.065173        1.037444        0.414084

                                    max             1.040318        1.730278        1.565701        1.301174

You can also apply the majority of NumPy universal functions to DataFrame objects:

In  [ 31 ]: np.sqrt(df)

Out[31]:                                                                        No1                         No2                         No3                         No4

                                    2015-01-31                          NaN     1.032072        0.270935        1.140690

                                    2015-02-28                          NaN                         NaN     0.635449                            NaN

                                    2015-03-31                          NaN                         NaN     1.018550                            NaN

                                    2015-04-30                          NaN     1.315400        0.645844        0.429045

                                    2015-05-31                          NaN                         NaN     0.314131                            NaN

                                    2015-06-30                          NaN     1.183985        1.107756        0.989641

                                    2015-07-31      0.677376                            NaN     1.251280                            NaN

                                    2015-08-31                          NaN                         NaN                         NaN                         NaN

                                    2015-09-30      1.019960                            NaN     0.886672        0.643494

NUMPY UNIVERSAL FUNCTIONS

In general, you can apply NumPy universal functions to pandas DataFrame objects whenever they could be applied

to an ndarray object containing the same data.

pandas is quite error tolerant, in the sense that it captures errors and just puts a NaN value

where the respective mathematical operation fails. Not only this, but as briefly shown

already, you can also work with such incomplete data sets as if they were complete in a

number of cases:

In  [ 32 ]: np.sqrt(df).sum()

Out[32]:    No1             1.697335

                                    No2             3.531458

                                    No3             6.130617

                                    No4             3.202870

                                    dtype:  float64

In such cases, pandas just leaves out the NaN values and only works with the other

available values. Plotting of data is also only one line of code away in general (cf.

Figure 6-1):

In  [ 33 ]: %matplotlib inline

df.cumsum().plot(lw=2.0)