我編寫了一個(gè)遞歸函數(shù)來增強(qiáng) pandas.DataFrame.describe。它將峰度和偏斜添加為行。它還創(chuàng)建了第二個(gè)描述表來轉(zhuǎn)置第一個(gè)描述表，以便您獲得匯總統(tǒng)計(jì)數(shù)據(jù)的摘要。它工作得很好，只是我不喜歡編寫具有多個(gè)出口的函數(shù)。我嘗試用一個(gè) return 語(yǔ)句編寫它（請(qǐng)參閱注釋掉的部分），但它會(huì)在一個(gè)表中創(chuàng)建兩個(gè)轉(zhuǎn)置表。正確，但太多了。def get_better_desc(df, recursions: int = 1): '''Adds kurtosis and skew to pandas.DataFrame.describe output. And, creates second transposed version of this table called on itself for summary stats of summary stats. Parameters: df: pandas.DataFrame, or Series but its super_desc isn't so meaningful. recursions: integer number of times to apply recursively to create super_desc. Default value of 1 is all that is necessary. Returns: better_desc: pandas.Dataframe (or Series) with kurtosis and skew added. super_desc: pandas.DataFrame (or Series) of better_desc transposed and made into a better_desc itself.''' kurt = df.kurtosis() kurt.name = 'kurt' skew = df.skew() skew.name = 'skew' better_desc = df.describe().append([kurt, skew]) if recursions > 0: super_desc = get_better_desc(better_desc.transpose(), recursions=(recursions - 1)) return better_desc, super_desc else: return better_desc# if recursions > 0:# super_desc = get_better_desc(better_desc.transpose(),# recursions=(recursions - 1))# else:# super_desc = better_desc # return better_desc, super_desc
1 回答
陪伴而非守候
TA貢獻(xiàn)1757條經(jīng)驗(yàn) 獲得超8個(gè)贊
該函數(shù)返回兩個(gè)數(shù)據(jù)幀，如果您只將其分配給一個(gè)變量，而不是忽略第二個(gè)返回值，它會(huì)將兩者分配給該變量。在遞歸調(diào)用中，您只需指定第一個(gè)表，即 returns 列表中的第一個(gè)元素[0]。
...
if recursions > 0:
super_desc = get_better_desc(better_desc.transpose(),
recursions=(recursions - 1))[0] #<=
else:
super_desc = better_desc
return better_desc, super_desc
添加回答
舉報(bào)
0/150
提交
取消