alignment

[Ingrid Schockaert]

Dear all,

A simple question: is there, technically speaking, a limit to the number of dimensions in an alignment/absolute alignment file?

regards,

Ingrid

[Gaëtan de Menten]

Hello Ingrid,

 

There is no theoretical limit but in practice you will be quickly limited by your computer memory and computing time. Given that all combinations of the dimensions must exist in the alignment array, the size of the array grows quickly. As an example, even if you have only two different possibilities per dimension and you have 32 dimensions, the array will have 2 ^ 32 elements (approx. 4.3 billion), which would take approximately 32 Gb of memory.

 

Hope it helps,

Gaëtan

Disclaimer: This e-mail may contain confidential information which is intended only for the use of the recipient(s) named above.
If you have received this communication in error, please notify the sender immediately and delete this e-mail from your system.
Please note that e-mail messages cannot be considered as official information from the Federal Planning Bureau.

[Ingrid Schockaert]

Dear Gaetan,

 

Thank you for your answer. In terms of memory or speed, my model has not encountered any problems. In terms of validity of the results, it's however another question....

 

I would like to try the following:

At the household level:

- num_persons: persons.count

- is_candidate: align (“probabilities of emigration by municipality.csv”, 70)

- int_migration_targets: load('int_migration_targets.csv', type=int) (int_migration_targets is an alignment file with the number of emigrating individuals by household position, age and gender)

- migrant_pool: align_abs(num_persons, int_migration_targets, filter=is_candidate, link=persons)

 

Will the “is_candidate procedure” insure the distribution of the emigrations over the municipalities (if we assume that the distribution by age, gender, housholdposition is the same over all municipalities)?

thanks in advance,

Ingrid