Formatting Issues with Arviz Posterior Plots

Questions v5
1

Hello,

I am fairly new to pymc and Arviz, so pardon me if my question is off-base. I have performed an inference and used arviz.plot_posterior to visualise the posterior distribution. However, I am noticing a few formatting issues. Specifically, the bounds of the HDI overlap between the different plots. Furthermore, I am having trouble showing the results in scientific notation. Similar issues extend to other arviz plots as well

I have attached a copy of a sample posterior plot, and the code I use to generate the plots

Posterior_Plot.pdf (143.4 KB)

def posterior_plot_az(directory, trace, variable_names, reference_value=None):
	
	# Posterior Plot
	az.style.use(["default","arviz-viridish"])

	posterior_params = az.plot_posterior(trace, hdi_prob=0.99, round_to=4, var_names=variable_names)

	for ax in posterior_params.flatten():
		ax.ticklabel_format(axis='x',style='sci') #attempt to sci notation
		for ticks_fix in range(len(variable_names)): # attempted fix
			[ticks.set_rotation(45) for ticks in ax.figure.axes[ticks_fix].get_xticklabels()]
	plt.savefig(directory+'/'+'Posterior_Plot'+'.svg',format='svg')
	plt.close('all')

I am currently using arviz version 0.12.1. Any help would be great!

2

You can try using plt.rcParams['figure.constrained_layout.use'] = True right after setting the style or using plt.tight_layout() before saving the figure.

I don’t know how high is your ess value, but I also think you might be showing too many digits for which you don’t really have enough precision to know.

I am unsure right now about using scientific notation or even if this is possible.