关于bootstrap方法，最通俗的理解

ygqalone 发表于 2013-3-17 15:16
我我研究的是农民工问题 。。。

呵呵我研究的是农村贫困问题。

moonriverluo 发表于 2013-3-17 17:06
呵呵我研究的是农村贫困问题。

newhotter 发表于 2013-3-16 20:28
就是瞎弄，不可靠。这就是理解。。

大实话

我研究的也是农民工问题，想问下楼主这个问题解决没有？

花开院落 发表于 2013-4-18 08:50
我研究的也是农民工问题，想问下楼主这个问题解决没有？

没呢

The most simple bootstrapping is just drawing samples repeatedly from your observations.

Say you have n observations. You draw k samples (k<=n, if k=n then it is full sample bootstrapping) each time. You would need to put the samples back to the observations after each drawing.

If your objective is mean value, then each draw gives you one mean. The entire bootstrapping (say 1000 draws) gives you 1000 means, which gives you a distribution of means. You can use the distribution for statistical inferences.

You can easily implement simple bootstrapping in Matlab using 'bootstrp' (or write a program yourself. it basically uses 'rand'). For higher level, you might need to write a program.