VPC plot with observations

1

Hi,
I was trying to plot VPC for my estimation, but in the VPC Plot, there are no observations; only the bandwidth is visible.

This is the code I used to run the VPC.

vpc1=vpc(est53; 
        samples=500,
        observations=[keys(est53.data[1].observations)[1]],
        quantiles=(0.1,0.5,0.9),
        level=0.95,
        ensemblealg=EnsembleSerial(),
        bandwidth=60,
        smooth=true)

final_vpc=vpc_plot(vpc1)

final model vpc
final model vpc999×583 47.6 KB

2

Lately there have been several research reports with Visual Predictive Check (VPC). Can anybody provide a layman intro to what these plots are allowing the analyst diagnose?
Bob

3

You would have to passobservations=true to the plotting function.

final_vpc=vpc_plot(vpc1; observations=true)

You can see the options for the plot by writing ?vpc_plot in the REPL (terminal).

4

It’s an interesting question, but also a bit off-topic in terms of the actual question being asked. Could you open a separate question at https://discourse.pumas.ai/c/diagnostics/vpc/. I’m sure someone will be happy to discuss the merits/pros/cons of VPCs or at least provide some references.

Best,
Patrick

5

I apologize @patrick.

6

No worries, I think it’s an interesting discussion. If you start a new tthread we can discuss there!

7

Thank you, Patrick. It works.

1 Like