With Power Pack DataExtensions comes a Historical Data Provider called "Random".
The Help promises some settings which *do not appear* when I create such a data set.
However, if I analyse the data generated I'd expect
* constant volatility
* constant Skewness
* constant Kurtosis
Instead I find:
This looks much more realistic, of course but sparks my curiosity:
How works the random data generator internally?
The Help promises some settings which *do not appear* when I create such a data set.
However, if I analyse the data generated I'd expect
* constant volatility
* constant Skewness
* constant Kurtosis
Instead I find:
This looks much more realistic, of course but sparks my curiosity:
How works the random data generator internally?
Rename
QUOTE:
The Help promises some settings which *do not appear* when I create such a data set.
No, it doesn't. The provider is not configurable and the Help doesn't promise otherwise.
QUOTE:
How works the random data generator internally?
Please see "How it works" in the Help > Extensions > Data Extensions > Random. It's documented there.
Help->Extensions->Data Extensions -> Random -> How it Works just says:
So ok, I misread "selects a randomization method" as an user settable option.
But either Simple random data generator nor Random walk generator are explained in any depth.
In my book an artificial random walk is just the accumulated sum of gaussian changes. Such data has constant volatility, zero skewness and constant kurtosis (if any).
The data produced is clearly different form this idea (see above).
The basic statistics change all the time and there are way too many huge gaps and large candles...
If forced to guessing I'd say its a bootstrap method of existing real world time series...
... so please, provide a few details to calm my brain...
... and to decide better if this data is useful for a "blind test."
QUOTE:
Each time a dummy history is built, provider randomly selects a randomization method: Simple random data generator Random walk generator. The choice for market bias (i.e. bull, bear or neutral market) is again made randomly.
So ok, I misread "selects a randomization method" as an user settable option.
But either Simple random data generator nor Random walk generator are explained in any depth.
In my book an artificial random walk is just the accumulated sum of gaussian changes. Such data has constant volatility, zero skewness and constant kurtosis (if any).
The data produced is clearly different form this idea (see above).
The basic statistics change all the time and there are way too many huge gaps and large candles...
If forced to guessing I'd say its a bootstrap method of existing real world time series...
... so please, provide a few details to calm my brain...
... and to decide better if this data is useful for a "blind test."
And indeed, the data from Random Data Source can have another characteristics:
This example is much loser to the "Random Walk generator" I'd expect:
* (nearly) zero Skewness
* (nearly) constant Kurtosis of 1.0
* (nearly) constant volatility
* no large gaps and no large candles
This example is much loser to the "Random Walk generator" I'd expect:
* (nearly) zero Skewness
* (nearly) constant Kurtosis of 1.0
* (nearly) constant volatility
* no large gaps and no large candles
Your Response
Post
Edit Post
Login is required