Top ten trolls #121

hyunjimoon · 2022-12-09T12:58:26Z

hyunjimoon
Dec 9, 2022
Maintainer

Top 10% (S = 11, 19) consumes 97% resource (fitting time): (1h 40m 33s + 1h 10m 42s) / (1h 40m 33s + 1h 10m 42s + 20 sec * 18) = 10275 / (10275 + 360) = .97. .97 is dominance index.

inferencedata link for reproducibility
among sbc.nc, S = 14 forms challenging geometry in general and S = 18 has one challenging corner which sometimes make chain =3 stuck

Generally fast (every S except 14, 19)

20 second fitting

05:25:03 - cmdstanpy - INFO - CmdStan start processing
chain 1 |          | 00:00 Status
chain 2 |          | 00:00 Status

chain 3 |          | 00:00 Status


chain 1 |▊         | 00:00 Status


chain 4 |▊         | 00:00 Status

chain 3 |▊         | 00:00 Status
chain 2 |▊         | 00:00 Status
chain 1 |█▋        | 00:04 Iteration:    1 / 1050 [  0%]  (Warmup)

chain 3 |█▋        | 00:04 Iteration:    1 / 1050 [  0%]  (Warmup)


chain 4 |█▋        | 00:04 Iteration:    1 / 1050 [  0%]  (Warmup)
chain 1 |██▌       | 00:06 Iteration:  100 / 1050 [  9%]  (Warmup)

chain 3 |██▌       | 00:06 Iteration:  100 / 1050 [  9%]  (Warmup)


chain 4 |██▌       | 00:07 Iteration:  100 / 1050 [  9%]  (Warmup)
chain 1 |███▎      | 00:08 Iteration:  200 / 1050 [ 19%]  (Warmup)

chain 3 |███▎      | 00:08 Iteration:  200 / 1050 [ 19%]  (Warmup)


chain 4 |███▎      | 00:09 Iteration:  200 / 1050 [ 19%]  (Warmup)
chain 1 |████▏     | 00:09 Iteration:  300 / 1050 [ 28%]  (Warmup)

chain 3 |████▏     | 00:10 Iteration:  300 / 1050 [ 28%]  (Warmup)


chain 4 |████▏     | 00:10 Iteration:  300 / 1050 [ 28%]  (Warmup)
chain 1 |█████     | 00:11 Iteration:  400 / 1050 [ 38%]  (Warmup)

chain 3 |█████     | 00:11 Iteration:  400 / 1050 [ 38%]  (Warmup)


chain 4 |█████     | 00:11 Iteration:  400 / 1050 [ 38%]  (Warmup)
chain 1 |█████▊    | 00:13 Iteration:  500 / 1050 [ 47%]  (Warmup)

chain 3 |█████▊    | 00:13 Iteration:  500 / 1050 [ 47%]  (Warmup)


chain 4 |█████▊    | 00:13 Iteration:  500 / 1050 [ 47%]  (Warmup)
chain 1 |██████▋   | 00:14 Iteration:  600 / 1050 [ 57%]  (Warmup)

chain 3 |██████▋   | 00:14 Iteration:  600 / 1050 [ 57%]  (Warmup)


chain 4 |██████▋   | 00:14 Iteration:  600 / 1050 [ 57%]  (Warmup)
chain 1 |███████▌  | 00:15 Iteration:  700 / 1050 [ 66%]  (Warmup)

chain 3 |███████▌  | 00:15 Iteration:  700 / 1050 [ 66%]  (Warmup)


chain 4 |███████▌  | 00:16 Iteration:  700 / 1050 [ 66%]  (Warmup)
chain 1 |████████▎ | 00:16 Iteration:  800 / 1050 [ 76%]  (Warmup)

chain 3 |████████▎ | 00:17 Iteration:  800 / 1050 [ 76%]  (Warmup)


chain 4 |████████▎ | 00:17 Iteration:  800 / 1050 [ 76%]  (Warmup)
chain 1 |█████████▏| 00:18 Iteration:  900 / 1050 [ 85%]  (Warmup)


chain 4 |█████████▏| 00:18 Iteration:  900 / 1050 [ 85%]  (Warmup)

chain 3 |█████████▏| 00:18 Iteration:  900 / 1050 [ 85%]  (Warmup)
chain 1 |          | 00:19 Iteration: 1001 / 1050 [ 95%]  (Sampling)


chain 4 |          | 00:19 Iteration: 1001 / 1050 [ 95%]  (Sampling)

chain 1 |██████████| 00:19 Sampling completed                       
chain 2 |██████████| 00:19 Sampling completed                       
chain 3 |██████████| 00:19 Sampling completed                       
chain 4 |██████████| 00:19 Sampling completed                       
05:25:23 - cmdstanpy - INFO - CmdStan done processing.

S-specific `S=11, M_c=3`

2 hour fitting log (1 out of 20)

05:26:02 - cmdstanpy - INFO - CmdStan start processing
chain 1 |          | 00:00 Status
chain 2 |          | 00:00 Status

chain 3 |          | 00:00 Status


chain 4 |          | 00:00 Status
chain 1 |▊         | 00:00 Status

chain 3 |▊         | 00:00 Status


chain 4 |▊         | 00:00 Status

chain 3 |█▋        | 00:04 Iteration:    1 / 1050 [  0%]  (Warmup)


chain 4 |█▋        | 00:05 Iteration:    1 / 1050 [  0%]  (Warmup)
chain 2 |█▋        | 00:05 Iteration:    1 / 1050 [  0%]  (Warmup)

chain 3 |██▌       | 00:07 Iteration:  100 / 1050 [  9%]  (Warmup)
chain 2 |██▌       | 00:07 Iteration:  100 / 1050 [  9%]  (Warmup)


chain 4 |██▌       | 00:07 Iteration:  100 / 1050 [  9%]  (Warmup)

chain 3 |███▎      | 00:09 Iteration:  200 / 1050 [ 19%]  (Warmup)
chain 2 |███▎      | 00:09 Iteration:  200 / 1050 [ 19%]  (Warmup)


chain 4 |███▎      | 00:10 Iteration:  200 / 1050 [ 19%]  (Warmup)

chain 3 |████▏     | 00:11 Iteration:  300 / 1050 [ 28%]  (Warmup)


chain 4 |████▏     | 00:11 Iteration:  300 / 1050 [ 28%]  (Warmup)
chain 2 |████▏     | 00:11 Iteration:  300 / 1050 [ 28%]  (Warmup)

chain 3 |█████     | 00:13 Iteration:  400 / 1050 [ 38%]  (Warmup)
chain 2 |█████     | 00:14 Iteration:  400 / 1050 [ 38%]  (Warmup)


chain 4 |█████     | 00:14 Iteration:  400 / 1050 [ 38%]  (Warmup)

chain 3 |█████▊    | 00:16 Iteration:  500 / 1050 [ 47%]  (Warmup)


chain 4 |█████▊    | 00:16 Iteration:  500 / 1050 [ 47%]  (Warmup)
chain 2 |█████▊    | 00:16 Iteration:  500 / 1050 [ 47%]  (Warmup)

chain 3 |██████▋   | 00:17 Iteration:  600 / 1050 [ 57%]  (Warmup)


chain 4 |██████▋   | 00:18 Iteration:  600 / 1050 [ 57%]  (Warmup)
chain 2 |██████▋   | 00:18 Iteration:  600 / 1050 [ 57%]  (Warmup)

chain 3 |███████▌  | 00:19 Iteration:  700 / 1050 [ 66%]  (Warmup)


chain 4 |███████▌  | 00:19 Iteration:  700 / 1050 [ 66%]  (Warmup)
chain 2 |███████▌  | 00:20 Iteration:  700 / 1050 [ 66%]  (Warmup)

chain 3 |████████▎ | 00:21 Iteration:  800 / 1050 [ 76%]  (Warmup)


chain 4 |████████▎ | 00:21 Iteration:  800 / 1050 [ 76%]  (Warmup)
chain 2 |████████▎ | 00:21 Iteration:  800 / 1050 [ 76%]  (Warmup)

chain 3 |█████████▏| 00:23 Iteration:  900 / 1050 [ 85%]  (Warmup)


chain 4 |█████████▏| 00:23 Iteration:  900 / 1050 [ 85%]  (Warmup)
chain 2 |█████████▏| 00:23 Iteration:  900 / 1050 [ 85%]  (Warmup)

chain 3 |          | 00:23 Iteration: 1001 / 1050 [ 95%]  (Sampling)


chain 4 |          | 00:24 Iteration: 1001 / 1050 [ 95%]  (Sampling)
chain 1 |██████████| 1:40:33 Sampling completed                       
chain 2 |██████████| 1:40:33 Sampling completed                     
chain 3 |██████████| 1:40:33 Sampling completed                     
chain 4 |██████████| 1:40:33 Sampling completed                     
07:06:36 - cmdstanpy - INFO - CmdStan done processing.

	Exception: normal_lpdf: Location parameter[1] is nan, but must be finite! (in '/Users/hyunjimoon/Dropbox/stanify/vignette/stan_files/Inven_S20N10R2_M200/data2draws.stan', line 73, column 8 to column 97)

	Exception: normal_lpdf: Location parameter[1] is nan, but must be finite! (in '/Users/hyunjimoon/Dropbox/stanify/vignette/stan_files/Inven_S20N10R2_M200/data2draws.stan', line 73, column 8 to column 97)

M_c-specific `S=19, M_c=3`

07:10:37 - cmdstanpy - INFO - CmdStan start processing
chain 1 |          | 00:00 Status
chain 2 |          | 00:00 Status

chain 3 |          | 00:00 Status


chain 4 |          | 00:00 Status
chain 2 |▊         | 00:00 Status

chain 3 |▊         | 00:00 Status


chain 1 |█▋        | 00:03 Iteration:    1 / 1050 [  0%]  (Warmup)
chain 2 |█▋        | 00:03 Iteration:    1 / 1050 [  0%]  (Warmup)


chain 1 |██▌       | 00:06 Iteration:  100 / 1050 [  9%]  (Warmup)
chain 2 |██▌       | 00:06 Iteration:  100 / 1050 [  9%]  (Warmup)


chain 1 |███▎      | 00:08 Iteration:  200 / 1050 [ 19%]  (Warmup)
chain 2 |███▎      | 00:08 Iteration:  200 / 1050 [ 19%]  (Warmup)


chain 1 |████▏     | 00:09 Iteration:  300 / 1050 [ 28%]  (Warmup)
chain 2 |████▏     | 00:09 Iteration:  300 / 1050 [ 28%]  (Warmup)


chain 1 |█████     | 00:11 Iteration:  400 / 1050 [ 38%]  (Warmup)
chain 2 |█████     | 00:11 Iteration:  400 / 1050 [ 38%]  (Warmup)


chain 1 |█████▊    | 00:12 Iteration:  500 / 1050 [ 47%]  (Warmup)


chain 4 |█████▊    | 00:12 Iteration:  500 / 1050 [ 47%]  (Warmup)
chain 1 |██████▋   | 00:14 Iteration:  600 / 1050 [ 57%]  (Warmup)


chain 4 |██████▋   | 00:14 Iteration:  600 / 1050 [ 57%]  (Warmup)
chain 1 |███████▌  | 00:15 Iteration:  700 / 1050 [ 66%]  (Warmup)


chain 4 |███████▌  | 00:15 Iteration:  700 / 1050 [ 66%]  (Warmup)
chain 1 |████████▎ | 00:16 Iteration:  800 / 1050 [ 76%]  (Warmup)


chain 4 |████████▎ | 00:17 Iteration:  800 / 1050 [ 76%]  (Warmup)
chain 1 |█████████▏| 00:19 Iteration:  900 / 1050 [ 85%]  (Warmup)


chain 4 |█████████▏| 00:19 Iteration:  900 / 1050 [ 85%]  (Warmup)
chain 1 |          | 00:20 Iteration: 1001 / 1050 [ 95%]  (Sampling)


chain 4 |          | 00:20 Iteration: 1001 / 1050 [ 95%]  (Sampling)
chain 2 |          | 00:20 Iteration: 1001 / 1050 [ 95%]  (Sampling)

chain 3 |█▋        | 04:53 Iteration:    1 / 1050 [  0%]  (Warmup)

chain 3 |██▌       | 11:45 Iteration:  100 / 1050 [  9%]  (Warmup)

chain 3 |███▎      | 18:29 Iteration:  200 / 1050 [ 19%]  (Warmup)

chain 3 |████▏     | 25:19 Iteration:  300 / 1050 [ 28%]  (Warmup)

chain 3 |█████     | 32:14 Iteration:  400 / 1050 [ 38%]  (Warmup)

chain 3 |██████▋   | 45:57 Iteration:  600 / 1050 [ 57%]  (Warmup)

chain 3 |███████▌  | 52:45 Iteration:  700 / 1050 [ 66%]  (Warmup)

chain 3 |████████▎ | 59:42 Iteration:  800 / 1050 [ 76%]  (Warmup)

chain 3 |█████████▏| 1:07:00 Iteration:  900 / 1050 [ 85%]  (Warmup)

chain 3 |██████████| 1:07:04 Iteration: 1000 / 1050 [ 95%]  (Warmup)

chain 1 |██████████| 1:10:42 Sampling completed                     
                                                                                                                                                                                                                                                                                                                                
chain 2 |██████████| 1:10:42 Sampling completed                     
chain 3 |██████████| 1:10:42 Sampling completed                       
chain 4 |██████████| 1:10:42 Sampling completed   
08:21:19 - cmdstanpy - WARNING - Some chains may have failed to converge.
	Chain 3 had 50 iterations at max treedepth (100.0%)
	Use function "diagnose()" to see further information.

hyunjimoon · 2022-12-09T14:09:35Z

hyunjimoon
Dec 9, 2022
Maintainer Author

Hypothesis:

based on the produced plots, bimodal posterior took a large time which we can verify with .nc file

4 replies

hyunjimoon Dec 9, 2022
Maintainer Author

Non-identifiability identified from left plot (left is parameter value, right is traceplot) @tomfid

hyunjimoon Dec 10, 2022
Maintainer Author

With Tom and Jair, we discovered

from time to time bimodal posteriors, outliers always have their other half on the relatively higher value (so that their mean is calibrated around the mean)
location nan happens because inventory adjustment time is sampled near zero, making flow big, which explains location parameter must be a finite warning msg. (in vensim, some problematic datapoints are ignored which can cause signal-noise confusion)
Jair recommended parameterizing 1/adj_time, and Tom thought this should be done during vensim modeling for consistency (not in stanify)

hyunjimoon Dec 10, 2022
Maintainer Author

result of the fraction:
every fitting ends in 1s with good recovery quality (formal sbc is to be done)

tomfid Dec 10, 2022
Maintainer

Nice!

hyunjimoon · 2022-12-16T11:10:13Z

hyunjimoon
Dec 16, 2022
Maintainer Author

I wish to dig deeper into the phenomenon of installing a soft barrier on zero using inverse parameterization. Zero-avoiding in normal(2, 2 * .1) is equivalent to infinite-avoiding in normal(1/2, 1/2 *.1) (hyperbolic geometry has a similar spirit) is the key. Could we find an analogy in econ or policy where inverse parameterization regulation is more effective or efficient or satisfactory? Instead of regulating based on current wealth, expertise, and popularity classification, what if we regulate based on 1/wealth, 1/expertise, 1/populatin? Would tax system become more effective or fairer, making extreme more extreme?

0 replies

hyunjimoon · 2022-12-17T14:21:43Z

hyunjimoon
Dec 17, 2022
Maintainer Author

speed distribution for 2hier_s_asy_osc_S10N40Q2P3H0R1_M1000_ps0.1 is:

180, 240, 240, 540, 540, 240, 13,000, 14,000, 280, 200 (s)

I wonder whether increasing process noise would make this distribution less polarized.

0 replies

hyunjimoon · 2023-09-03T02:41:21Z

hyunjimoon
Sep 3, 2023
Maintainer Author

0 replies

Top ten trolls #121

Uh oh!

Uh oh!

hyunjimoon Dec 9, 2022 Maintainer

Generally fast (every S except 14, 19)

S-specific S=11, M_c=3

M_c-specific S=19, M_c=3

Replies: 4 comments · 4 replies

Uh oh!

Uh oh!

hyunjimoon Dec 9, 2022 Maintainer Author

Uh oh!

Uh oh!

hyunjimoon Dec 9, 2022 Maintainer Author

Uh oh!

hyunjimoon Dec 10, 2022 Maintainer Author

Uh oh!

hyunjimoon Dec 10, 2022 Maintainer Author

Uh oh!

tomfid Dec 10, 2022 Maintainer

Uh oh!

hyunjimoon Dec 16, 2022 Maintainer Author

Uh oh!

Uh oh!

hyunjimoon Dec 17, 2022 Maintainer Author

Uh oh!

hyunjimoon Sep 3, 2023 Maintainer Author

hyunjimoon
Dec 9, 2022
Maintainer

S-specific `S=11, M_c=3`

M_c-specific `S=19, M_c=3`

Replies: 4 comments 4 replies

hyunjimoon
Dec 9, 2022
Maintainer Author

hyunjimoon Dec 9, 2022
Maintainer Author

hyunjimoon Dec 10, 2022
Maintainer Author

hyunjimoon Dec 10, 2022
Maintainer Author

tomfid Dec 10, 2022
Maintainer

hyunjimoon
Dec 16, 2022
Maintainer Author

hyunjimoon
Dec 17, 2022
Maintainer Author

hyunjimoon
Sep 3, 2023
Maintainer Author