Reply With QuoteI dont understand this question. If you know that after a certain time t a drift was introduced, why would you need to find out exactly when it was introduced? It was introduced at t.
Hi guys,
I am working on the following problem related to random walks. I have a set of data as a simple random walk starting at t=0, after certain time t, a drift is introduced. I can see the different distributions for before and after the drift was introduced. How would you find when exactly the drift was introduced?
Any suggestions greatly appreciated!!
I dont understand this question. If you know that after a certain time t a drift was introduced, why would you need to find out exactly when it was introduced? It was introduced at t.
"Facts are stubborn things, but statistics are more pliable." Mark Twain
I think they're saying that it is known that at some point a drift was introduced but that the exact time isn't known.
A number of approaches could be taken here. Are you more comfortable with frequentist or Bayesian analysis? The thing to note though is that it's probably not possible to tell exactly when the drift started but you could probably come up with a reasonable set of plausible values for the time it started.
"His programming is malfunctioning. It begins! Get your weapons, he's going to become a killbot!!!" - bryangoodrich
One possibility is look at the data for some specific point that the mean began to change (which is imperfect at best). You can measure different intervals and find the first that white noise is not indicated, but this also won't be precise and will take a long time. I assume in both of these points there is no trend before the walk.
Another possibility is to determine theoretically when the walk began and then test this.
"Facts are stubborn things, but statistics are more pliable." Mark Twain
Posting Permissions
|
|
