b*******d 发帖数: 190 | 1 一个时间序列,怎么测试它是不是stationary的?我从网上查到可以做unit root
testing,但是这个只是测第一个AR coefficient是否小于1。除此之外,怎么测
variance和autocorrelation是不是恒定呢? |
b*****n 发帖数: 685 | 2 我想如果series够长的话,原则上你可以分成很多段,然后在每段中做bootstrap,再
用anova来test之 |
l*********s 发帖数: 5409 | 3 there are general statistical test for stationality, but they can break your
head if you are not math major.
【在 b*******d 的大作中提到】 : 一个时间序列,怎么测试它是不是stationary的?我从网上查到可以做unit root : testing,但是这个只是测第一个AR coefficient是否小于1。除此之外,怎么测 : variance和autocorrelation是不是恒定呢?
|
d******g 发帖数: 130 | 4 我记得我做过的empirical research中,用过plot raw data against time, 看是否有
明显的trend,variance 是否stablized.还可以plot ACF,看if acf varies across
time.R的car里面有box.cox.powers,可以依据box.cox transformation检测variance的
stablization情况,依据rank.可以看看相关的r doc on box.cox.powers.
【在 b*******d 的大作中提到】 : 一个时间序列,怎么测试它是不是stationary的?我从网上查到可以做unit root : testing,但是这个只是测第一个AR coefficient是否小于1。除此之外,怎么测 : variance和autocorrelation是不是恒定呢?
|
b*******d 发帖数: 190 | 5 do you know what they are called? Are there built in programs in R or Matlab
that can do it?
your
【在 l*********s 的大作中提到】 : there are general statistical test for stationality, but they can break your : head if you are not math major.
|
l*********s 发帖数: 5409 | 6 Unfortunately I don't. Our professor just alluded this in the class of time
series without further details.
Matlab
【在 b*******d 的大作中提到】 : do you know what they are called? Are there built in programs in R or Matlab : that can do it? : : your
|
b*******d 发帖数: 190 | 7 thx!
I think it's ironic that stationarity is such a big deal in time series
analysis but testing for it is so hard :)
time
【在 l*********s 的大作中提到】 : Unfortunately I don't. Our professor just alluded this in the class of time : series without further details. : : Matlab
|
l*********s 发帖数: 5409 | 8 Indeed, the gap between theory and practice is always very wide.
【在 b*******d 的大作中提到】 : thx! : I think it's ironic that stationarity is such a big deal in time series : analysis but testing for it is so hard :) : : time
|
b*******d 发帖数: 190 | 9 谢谢
我现在的问题是我的data肯定是有trend的,我要编一个程序来detrend它直到
stationary,所以需要一些objective的test而不是肉眼看
我看了一下,R的car里面没有box.cox.powers这个命令啊,有一个bcpower,不知道你
说的是不是这个
【在 d******g 的大作中提到】 : 我记得我做过的empirical research中,用过plot raw data against time, 看是否有 : 明显的trend,variance 是否stablized.还可以plot ACF,看if acf varies across : time.R的car里面有box.cox.powers,可以依据box.cox transformation检测variance的 : stablization情况,依据rank.可以看看相关的r doc on box.cox.powers.
|