Move to x0 fixed for everything and rearrange all data/init_params files

[jlperla]

• [x] Make sure nothing is missing from this list!
• [x] Verify that the x0 isn't drawn for any of the main experiments
  • remove comments if just commented out
• [x] Verify that the calculate_ergodic_distribution=false is globally turned off. No need to calculate it for any of the examples after this?
• [x] Regenerate all data fixing x0
  • I moved the generate_simulated_data_rbc_frequentist.jl inside the deps.
  • Huge numbers of files, but not much space. Right now it isn't committed but eventually we can.
• [x] Do new burnins/initial conditions with the fixed x0 for all.
• [x] What do we need to do on the dynare? I think it is just when we rerun it?
• [x] While we are at it, lets change our naming convention slightly to include the sample length
  • rbc_1.csv=> rbc_1_100.csv
  • etc. Then we have rbc_1_100.csv as the baseline and rbc_1_500.csv for the longer sample, etc.
  • Then lets rename the initial conditions and put in a subfolder:
  • rbc_1_joint_burnin.csv => init_params/rbc_1_joint_init_params_100.csv
• [x] Get rid of all data files and initial conditions we are not using for the experiments.
• [x] For the deps look for the text x_ergodic_var and make sure to set to zeros.
• [x] Anything that says high vs. low etc. in the data and init_params should be removed and we pick the one consisetnt with the noise in the experiment. i.e. the default.
• [x] When we regenerate the burnin files for RBC with the joint case I think we were careful (hence the _dynare name?) to have the deep parametesr start at the same point as the dynare code to make more directly comparable? We should verify that. i.e. maybe we just want to copy over the deep parameter initial conditions over top of the file?

[jlperla]

See #145

[jlperla]

For the deps look for the text x_ergodic_var and make sure to set to zeros.

[jlperla]

Anything that says high vs. low etc. in the data and init_params should be removed and we pick the one consisetnt with the noise in the experiment. i.e. the default.

Good chance to make sure things are consistent with the observational noise in the simulation of data and the default parameters. SGU kalman vs. joint was the big question.