多个时间序列的描述性统计量，求助！！

最近做个数据分析，有800个时间序列，现在有点迷糊，这个数据集的描述性统计量该怎么做？一个时间序列好办，但这么多，该咋做呢？有大家帮忙！！！另外,做出后能否做成个表,好转化为PDF或者其他图片格式?

mixed effect model

If you use R, then it's easy, try package"coda". Lets say you have a time series X(T,N), N=800 in your case,
then just write following code:

# 1. define as a mcmc object
mc.X=mcmc(X)

# 2. some descriptive summary statistics
summary(X)  # or partily of X, i.e., X[,1:5]

# 3. some density plot
plot(X)  # or partily of X, i.e., X[,1:5]

# 4. if you want to learn more about this package
help(package=coda)

Sorry for english, I only hv english type in my office computer, but I can see Chinese :)

Bayesian.var 发表于 2011-10-16 23:42
If you use R, then it's easy, try package"coda". Lets say you have a time series X(T,N), N=800 in yo ...

你好！我一直比较困惑的是对这么多时间序列最后的描述统计量的形式：如果按照每个时间序列，则描述的是单个时间序列的特征，如果按照某一横截面（某特定时间点或区间）则描述的又是横截面特征。现在，你给出的这个方法我也尝试了一下，但结果还是一个个时间序列的特征。我曾经设想：先将所有时间序列的统计量得出，再针对各个统计量再进行描述统计，比如，N个序列的均值，再描述均值：最大、最小、均值标准差等等，但未见任何人做过，所以困惑——但我觉得这样是对数据更好地描述！ 哦，对了，你的第二个语句：summary(X)是不是应该是summary(mc.X)，否则又回到了初始数据直接描述了！ anyway，thanks!

是呀，说得对

elephann 发表于 2011-10-18 19:01
你好！我一直比较困惑的是对这么多时间序列最后的描述统计量的形式：如果按照每个时间序列，则描述的是 ...

Hi,
Thanks for your correction, and I am sorry that that I was misunderstanding with your question before. But now my question arises, why do you need to give a kind of "descriptive" statistics about so many series at one time, and that should also makes the others so easily to understand, and assume that they don't need a formula or further inference to help understanding. For instance, if we have a series of 800 funds, I think it is really hard to give a such "descriptive" statistics that can pool information together based on all, or maybe it's not a good idea to pool them together just because we want to describe it shortly, or maybe you are refering to the stock index (Dow Jones) or the economic index, or dynamic factors..., if you create such an index as you metioned (Actually I am really looking forward to it :)), people may need instructions to understand your index, and to know how your index describes the series.
Anyway, it is a good question when we face such an "embrassment of riches". If I were you, I may try to think something that is "representative", like Top 5, top 10 series, then to say something "descriptive"; or even better, if I know how the series can be grouped or blocked, like I get the data from different sections of economy, or a block of parameters (L=(l1,l2,....)'), then I will pick up one or some out and describe with usual statistics.
            

看看

这算是纵向数据吗？

kankan

学习