media coefficient posterior distribution comes out too sparse

[ghost]

First, I very much appreciate all the efforts put into this work. It is super informative and well explained.

I ran the provided notebook as is but I do not get the same results as seen in the medium post.

When I plot media coefficients, their distribution is substantially sparse compared to what was demonstrated.

Anyone also experiencing the same?

The only difference I have from the original is my run environment; I am using PyStan 3 instead of 2.

[ghost]

I have resolved the issue myself now. I found that the code below appends posterior samples column-wise instead of row-wise. After switching i and j assignments, I am able to get the same plots as in the original medium post.

# plot media coefficients' distributions
# red line: mean, green line: median
beta_media = {}
for i in range(len(mmm['media_vars'])):
    md = mmm['media_vars'][i]
    betas = []
    for j in range(len(mmm['beta_list'])):
        betas.append(mmm['beta_list'][j][i])
    beta_media[md] = np.array(betas)

f = plt.figure(figsize=(18,15))
for i in range(len(mmm['media_vars'])):
    ax = f.add_subplot(5,3,i+1)
    md = mmm['media_vars'][i]
    x = beta_media[md]
    mean_x = x.mean()
    median_x = np.median(x)
    ax = sns.distplot(x)
    ax.axvline(mean_x, color='r', linestyle='-')
    ax.axvline(median_x, color='g', linestyle='-')
    ax.set_title(md)