Hi. Is 12345 the only seed to generate weibull random numbers with the function wblrnd?

When using Entity generation block in simulink, the default seed is 12345, provided for intergeneration time action (by MATLAB ACTION) for random number with weibull or exponential distribution.

Why 12345 instead of a seed 0 or 1? What justifies this choice in simulink? Is-it possible to use others seeds with weibull distribution for different replications in simulink? Thanks a lot.

Answer 1

[Walter Roberson]

Use any seed you want that is an integer between 1 and 4294967295 (inclusive)

rng seed 0 is not permitted.

[Guy YAKANA]

Thanks. Do you have any information about different seeds for the 5 number random generators? Thanks again.

[Walter Roberson]

@Guy YAKANA

Not sure what you mean? There are 8 different random number generators

https://www.mathworks.com/help/matlab/math/creating-and-controlling-a-random-number-stream.html

As far as I know, none of the seeds have special properties.

[Guy YAKANA]

To be clear, is 12345 a particular seed for a 5 random generator. This is what i read in a scientific article dealing with simulation : "Runs used randomly assigned treatments by choosing at random different seeds for the 5 ranndom number generators".

So i looking in MATLAB to see where is the link between the seed 12345 provided by simulink, and what i read in that article.

[Walter Roberson]

There is really nothing technically special about the choice of 12345 . The author had to choose some value, and "12345" is common for people to choose: long enough that people to feel they have done some "work", but not so long that people feel as if they are "trying too hard".

If the authors had used (for example) 74913 then people reading the code might well wonder why that value, what special properties that seed has. But when 12345 is used, people know that it was just an arbitrary number with no special meaning.

[Guy YAKANA]

So, can you share what the text i quoted means to you : "Runs used randomly assigned treatments by choosing at random different seeds for the 5 ranndom number generators".

[Walter Roberson]

That appears to be an extract from https://ur.booksc.eu/book/37589629/02946d

Reading the paper... I do not know. L27 table and 4 replications = 27*4 = 108 that was used... so there is no obvious link between "five random number generators" and the number of replications.

[Guy YAKANA]

Yes, you're right.

[Guy YAKANA]

Is the number of replications equal to the number of seeds for simulation?

[Walter Roberson]

No. That is why I multiplied out the 27*4 = 108 -- 27 table entries, 4 replications = 108 total cases that they explicitly state. sometimes when people talk about "4 replications" they mean "1 original and 4 more as well", and if that were the case then the total would be divisible by 5. But 108 is not divisible by 5, so 4 is the total number not "1 and 4 more".

So I do not currently understand why they talk about 5 random number generators.