I am reproducing the Ambulance-attended overdoses model part of overdose modelling methods draft.pdf. The model I refer is below.
For simplicity, I am only working with the overdose model part and call-outs part excluding intervention model first.
I have a few questions.
Where is the parameters we set ?
Previous section of this section deals with Time-series infectious disease model (TSIR).
The TSIR model has the parameters we wish to estimate parameters of interest (c_0, c_w, theta, alpha and p). We already knows the values of them. We generate a sample data from the parameters.
Then We set a prior for them, we fit the model, and obtain posterior sample values. The Figure 2 is about how well the credible intervals from posterior samples can cover the true values of parameters.
In Ambulance-attended overdoses model section that I am working on, I. cannot see the parameter values. For example, the mu and sigma for z_t and is not set. prior distribution of p_a seems to be not in the paper. Don't we need to set the parameter values and to give a distribution to the parameters?
I feel difficult to see the connection between the overdose model and call-outs.
From my naive understanding, the overdose model and call-outs model are separate (I mean they do not share any same parameter or variable...). So I am struggling what kinds of connection there can be. I believe Figure 6 demonstrate the effect of the size of n_a to the estimate of overdoses (O_t), and to the estimation of p_a by Figure 6 (a) and (b) respectively.
But how can they be combined and give us the result ? I would like to know how can we derive the result.
I hope you enjoying the black weekend.
I am reproducing the Ambulance-attended overdoses model part of overdose modelling methods draft.pdf. The model I refer is below.
For simplicity, I am only working with the overdose model part and call-outs part excluding intervention model first.
I have a few questions.
Previous section of this section deals with Time-series infectious disease model (TSIR). The TSIR model has the parameters we wish to estimate parameters of interest (c_0, c_w, theta, alpha and p). We already knows the values of them. We generate a sample data from the parameters. Then We set a prior for them, we fit the model, and obtain posterior sample values. The Figure 2 is about how well the credible intervals from posterior samples can cover the true values of parameters.
In Ambulance-attended overdoses model section that I am working on, I. cannot see the parameter values. For example, the mu and sigma for z_t and is not set. prior distribution of p_a seems to be not in the paper. Don't we need to set the parameter values and to give a distribution to the parameters?
I feel difficult to see the connection between the overdose model and call-outs. From my naive understanding, the overdose model and call-outs model are separate (I mean they do not share any same parameter or variable...). So I am struggling what kinds of connection there can be. I believe Figure 6 demonstrate the effect of the size of n_a to the estimate of overdoses (O_t), and to the estimation of p_a by Figure 6 (a) and (b) respectively. But how can they be combined and give us the result ? I would like to know how can we derive the result.
pymc3 error: Bad initial energy I generated Jupiter notebook for this question. https://github.com/aiod01/Bayesian-evidence-synthesis-/blob/master/questions2.ipynb
I am getting this error: ParallelSamplingError: Bad initial energy
I've been googled about this but it seems there are many causes of this problem. Could you help me to understand the issue here ?
Thanks in advance.