我正在使用Seaborn的tsplot
来查看时间序列(可用数据here):
Vent (m/s) year month date mois
7.9456207374248 1998 1 2012-01-01 00:00:59 1
7.49342837173539 1998 1 2012-01-01 01:00:59 1
7.49853611345398 1998 1 2012-01-01 02:00:59 1
7.61488603321423 1998 1 2012-01-01 03:00:59 1
7.70778826901725 1998 1 2012-01-01 04:00:59 1
7.78203058333749 1998 1 2012-01-01 05:00:59 1
8.00562302385017 1998 1 2012-01-01 06:00:59 1
8.14370922860093 1998 1 2012-01-01 07:00:59 1
8.28552955459094 1998 1 2012-01-01 08:00:59 1
8.60182223260882 1998 1 2012-01-01 09:00:59 1
9.14166286842826 1998 1 2012-01-01 10:00:59 1
sns.tsplot(value = 'Vent (m/s)', data = df, time ='mois', condition="year", unit = "date",
interpolate=False,
estimator = np.nanmean,
#ci=[68, 100],
)
这是输出:
我发现CI频段非常违反直觉:我期望CI显示每个月的标准错误,但CI值不同。实际上,df[df["month"] == 1]["Vent (m/s)"].std()
返回3.5529405282100215
而情节读数低于1.
我做错了吗?