Questions for var_names in sample_posterior_predictive

Questions
1

Dear all,

May I know is anyone has experience with setting var_names in sample posterior predictive?
I faced the error when I use it.

image
image976Ă—707 127 KB

I tried change my data from series to tuple, it give me an error of inmutable data cannot be hashed.

Please help. Thanks alot.

2

Can we see the contents of final_train?

3

Hi cluhmann,

image

Correct me if I am wrong.
We need to convert it into RV class?

Thanks

4

The var_names argument to posterior_predictive() should be a list of parameter names. See this notebook for some usages examples.

1 Like
5

thanks alot for your answer. I will take a look on that

6

Hello

I am trying to run the following code, where it keeps on giving “Keyerror: mu”. Though as can be seen that “mu” is declared as var_names. Need some suggestions to get rid of this error.

models = {1: first, 2: second, 3: third, 4: fourth} # dictionary of the models
samples = [
first_sample,
second_sample,
third_sample,
fourth_sample,
] # list the posterior samples
orders = [1, 2, 3, 4]

_, ax = plt.subplots()

for i, sample in zip(models.keys(), samples): # iterate through models and samples
with models[i]: # set the model
x = np.linspace(-3.1, 1.3, 100)
#pm.set_data({“x”: x})
ppc = pm.sample_posterior_predictive(sample, var_names=[“mu”])
ax.plot(x, ppc[“mu”].mean(0), label="order " + str(i))
sns.scatterplot(x=“s_taxRate”, y=“s_taxRevenue”, data=Laffer)
ax.set_title(“Weakly regularized models”);

az.summary(second_sample, var_names=[“b”])

7

mu should be in ppc.posterior_predictive

1 Like
8

Thank you @ricardoV94. Your suggestion solved my issue.