Errata

In [169]: points = np.arange(-5, 5, 0.01) # 100 equally spaced points

-> this will return 1000 equally spaced points, not 100

In the two examples for data type for the array that NumPy creates:
In [27]: arr1.dtype
Out[27]: dtype('float64')

In [28]: arr2.dtype
Out[28]: dtype('int64')

The output of the dtype for arr2 is not int64 but int32.

Note from the Author or Editor:
I will add a note that the output might be int32 on some platforms

Expression:
Axis to resample on; default `axis=0`

Suggestion for improvements:
Axis to resample on; default `axis="index"`

Reason:
This is not a mistake, but since the 3rd edition seems to unify the specification of axis in pandas with `"index"` and `"columns"` instead of numbers, the specification with numbers may surprise the reader a little.

Note from the Author or Editor:
I am fixing in text

Error:
`facecolor, edgecolor`
The color of the figure background outside of the subplots; `"w"` (white), by default.

Correct:
The color of the figure background outside of the subplots; default to `rcParams["savefig.facecolor"]` and `rcParams["savefig.edgecolor"]`, both of which default to `"auto"` (facecolor and edgecolor of the current figure).

Reason:
The default changed from matplotlib 3.3.

Note from the Author or Editor:
I'm removing the part about the default altogether since it's pretty in the weeds

Error:
You can save the active figure to file using the figure object’s savefig instance method.

Correct:
You can save the figure to file using the figure object’s savefig instance method.

Reason:
In the 2nd ed., the target of the operation was an active figure since the section described `plt.savefig`, but in the 3rd ed., since the `savefig` instance method of a figure object is described, I think the target of the operation does not need to be active.

Note from the Author or Editor:
I am removing "active" from the text

Error:
In [50]: ax.legend()

The `legend` method has several other choices for the location `loc` argument. See the docstring (with `ax.legend?`) for more information.

The `loc` legend option tells matplotlib where to place the plot. The default is `"best"`, which tries to choose a location that is most out of the way. To exclude one or more elements from the legend, pass no label or `label="_nolegend_"`.

Correct:
In [50]: ax.legend()

The `legend` method can take the `loc` option, which instructs matplotlib where to place the legend in the plot. The `loc` option defaults to `"best"`, which tries to choose a location that is most out of the way. To exclude one or more elements from the legend, pass no label or `label="_nolegend_"`. The `legend` method has several other choices for the `loc` argument. See the docstring (with `ax.legend?`) for more information.

Reason:
In the 2nd ed., the author passed `loc="best"` as the argument to `legend` in the code block 50, so readers could read the subsequent sentences under the assumption that the `loc` option could be passed. In the 3rd ed., the `loc` option is not passed to `legend` in the code block 50, so the explanation of the `loc` option seems abrupt.

Note from the Author or Editor:
I revised the text in this section

uniform appears two times in the table

Note from the Author or Editor:
fixing

the Python built-in range() function does not return a list but a generator

Note from the Author or Editor:
fixing

In Table 9-4, I believe the argument is "layout" (singular).

Note from the Author or Editor:
will fix

The line "In [151]: ..." appears to be superfluous---a holdover from the second edition.

Note from the Author or Editor:
will fix

It is *not* true that "if any value is not NA, then the result is NA." Apparently the default is to skip (exclude) NA values.

Note from the Author or Editor:
Yes, the language needs to be fixed to indicate that the result will be the sum of the non-NA values

The randomly generated array (as below) is inappropriate as an example, as the first column is in ascending order from the beginning. Therefore, although we want only the first column to be sorted, there is no change in the array before and after sorting, which makes it difficult to convey the intent.

It would be an appropriate example if it were generated with other parameters.

In [166]: arr = rng.standard_normal((3, 5))

In [167]: arr
Out[167]:
array([[-1.1956, 0.4691, -0.3598, 1.0359, 0.2267],
[-0.7448, -0.5931, -1.055 , -0.0683, 0.458 ],
[-0.07 , 0.1462, -0.9944, 1.1436, 0.5026]])

In [168]: arr[:, 0].sort() # Sort first column values in place

In [169]: arr
Out[169]:
array([[-1.1956, 0.4691, -0.3598, 1.0359, 0.2267],
[-0.7448, -0.5931, -1.055 , -0.0683, 0.458 ],
[-0.07 , 0.1462, -0.9944, 1.1436, 0.5026]])

Note from the Author or Editor:
I will improve the example to be more robust to random number generation

Error:

As you may recall from Ch 8: Data Wrangling: Join, Combine, and Reshape, pandas has some tools, in particular pandas.cut and pandas.qcut, for slicing data up into buckets with bins of your choosing, or by sample quantiles.

Correct:
As you may recall from Ch 7: Data Cleaning and Preparation, pandas has some tools, in particular pandas.cut and pandas.qcut, for slicing data up into buckets with bins of your choosing, or by sample quantiles.

Reason:
pandas.cut and pandas.qcut are discussed in Ch 7 Section 2, Discretization and Binning.

Note from the Author or Editor:
will fix

In the first sentence of the preface here:
wesmckinney.com/book/preface.html

it says:
"This Open Access web version of Python for Data Analysis 3rd Edition is now available as a companion to the print and digital editions. If you encounter any errata, please report them here."
The URL for error reporting is: www.oreilly.com/catalog/errata.csp?isbn=0636920023784

that is the wrong URL. The correct URL is: oreilly.com/catalog/0636920519829/errata

Note from the Author or Editor:
will fix