pandas滚动回归实现

python的pandas包的groupby似乎难以直接滚动回归。
原因似乎是rolling的结果是一个Series。
代码如下：

import pandas as pd

df = pd.DataFrame({"a": [i for i in range(100)], "b": [i ** 2 for i in range(100)]})

def tmp(x):
    print(x)
    return 1
s = df.rolling(3).apply(tmp)

解决方法：使用numpy的as_strided 方法
直接上代码

import pandas as pd
import numpy as np

from numpy.lib.stride_tricks import as_strided as stride


def test(df):
    model = np.polyfit(df["a"], df["b"], deg=1)
    return model[0], model[1]


class GetRolling(object):
    def __init__(self, df):
        self.df = df
        self.nrows, self.ncols = df.shape[0], df.shape[1]
        self.strides = df.values.strides

    def getGroup(self, rolling_nums):
        my_stride = stride(df.values, (self.nrows - rolling_nums + 1, rolling_nums, self.ncols),
                           (self.strides[0], self.strides[0], self.stri