Setting up
Multiple parameter vectors
If we look at .glurmo/settings.json
we will see a new change in addition to the one from the previous section:
...
"N" : "@[100, 200, 300]",
"beta_0" : "1",
"beta_1" : "@[1, 2, 3]",
...
Now, in addition to varying beta_1
, we’ll be varying the sample size N
as well.
Setting up
We can set up this simulation in exactly the same way as before. As always, ensure you are in the simulation directory and run:
$ glurmo -s .
You will see substantially more messages showing the progress of the setup, though this should not take too long. If you look at the contents of varied-lm-2
, you should now see three subdirectories: N_100
, N_200
, and N_300
. These are exactly analogous to the subdirectories that were set up in the previous example.
Let’s check the settings file in N_100
, which can be found in N_100/.glurmo/settings.json
. Specifically, let’s look at the entry for beta_1
:
...
"beta_1": "@[1, 2, 3]"
...
This again speaks to the recursive nature of glurmo
directories: each setting for N
is just a glurmo
simulation with N
set to a specific value, and aside from that all of the same parameter settings (including parameter vectors).
It is worth emphasizing that by changing a single line in our settings file (not including the simulation id), we went from having three simulation settings to nine. This is one of the really nice things about .glurmo
: it makes it easy to alter and scale up simulations as necessary.