Command: rdevdict::*_clipped

Description

Generate random numbers from an underlying distribution

but restricted to a certain interval.

For continuous distributions

low < random < high

will hold true i.e. numbers are restricted to the open interval (low high).

If the distribution itself is restricted to e.g. positiv numbers setting

low < 0 will still only return positive numbers. Clipping only excludes numbers

outside (low high).

For discrete distribtions numbers are limited to {low low+1 ... high} i.e.

in this case min and max are included.

Numbers are clipped by re-drawing any number outside (low high) until

a number in (log high) is drawn. The actual distribution of random

numbers drawn will be a distorted version of the underlying

distribution.

/low lower bound (default: -inf)

/high upper bound (default: +inf)

but restricted to a certain interval.

For continuous distributions

low < random < high

will hold true i.e. numbers are restricted to the open interval (low high).

If the distribution itself is restricted to e.g. positiv numbers setting

low < 0 will still only return positive numbers. Clipping only excludes numbers

outside (low high).

For discrete distribtions numbers are limited to {low low+1 ... high} i.e.

in this case min and max are included.

Numbers are clipped by re-drawing any number outside (low high) until

a number in (log high) is drawn. The actual distribution of random

numbers drawn will be a distorted version of the underlying

distribution.

Parameters

/low lower bound (default: -inf)

/high upper bound (default: +inf)

Author

Hans Ekkehard Plesser

- Clipped generators can be very inefficient if there is little probability

mass in (low high).

- For continuous distributions the probability of actually drawing

min or max is approximately 2^-62 ~ 2e-19 i.e. negligible. Therefore

one can only clip to open intervals.

- There are also *_clipped_to_boundary versions of the generators. These

return the respective boundary value if a value outside the interval is

drawn. These versions are mainly provided to allow reproduction of

publications that used this strategy.

Remarks

- Clipped generators can be very inefficient if there is little probability

mass in (low high).

- For continuous distributions the probability of actually drawing

min or max is approximately 2^-62 ~ 2e-19 i.e. negligible. Therefore

one can only clip to open intervals.

- There are also *_clipped_to_boundary versions of the generators. These

return the respective boundary value if a value outside the interval is

drawn. These versions are mainly provided to allow reproduction of

publications that used this strategy.

File

librandom/clipped_randomdev.h