Move to v4 for Rethinking 2 #194
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
View / edit / reply to this conversation on ReviewNB aloctavodia commented on 2022-06-02T16:59:54Z This need to fixed https://github.com/pymc-devs/pymc/issues/5443
see how you are overestimating the standard deviation |
|
View / edit / reply to this conversation on ReviewNB aloctavodia commented on 2022-06-02T16:59:55Z Line #2. sns.kdeplot(x[4, :].flatten(), bw_method=0.01, ax=ax[0]) alternatively, we can use az.plot_kde |
|
Issue pymc-devs/pymc#5443 seems to be over my pay grade, going to exclude Yes indeed, using |
Sorry for not being clear: This is the change needed in data = np.repeat((0, 1), (3, 6))
with pm.Model() as m:
p = pm.Uniform("p", 0, 1) # uniform priors
w = pm.Binomial("w", n=len(data), p=p, observed=data.sum()) # binomial likelihood
mean_q = pm.find_MAP()
p_value = m.rvs_to_values[p]
p_value.tag.transform = None
p_value.name = p.name
std_q = ((1 / pm.find_hessian(mean_q, vars=[p])) ** 0.5)[0]
# display summary of quadratic approximation
print(" Mean, Standard deviation\np {:.2}, {:.2}".format(mean_q["p"], std_q[0])) |
…nb with RNg update
|
View / edit / reply to this conversation on ReviewNB aloctavodia commented on 2022-06-03T14:57:08Z This cell also needs to be fixed, similar to the previous one.
|
Move to v2 for Rethinking 2
Starting with Statistical Rethinking 2 as this is the one I know the best.
Was using the script from pymc-example (
scripts/rerun.py)Some of the file got transformed automatically:
Rethinking_2/Chp_02.ipynbRethinking_2/Chp_03.ipynbRethinking_2/Chp_10.ipynbRethinking_2/End_of_chapter_problems/Chapter_2.ipynbRethinking_2/End_of_chapter_problems/Chapter_11.ipynbThis PR include #152 .
For
Rethinking_2/Chp_04.ipynb:return_inferencedata=Truefor thepm.sampleso using theaz.InferenceDatafor the the trace variables.trace['var']totrace.posterior('var']when possible or when it was expecting a array with all the sample in one array (rather than one array per chain), I usedaz.extract_datasetOther change needed was when a Panda Series was used to define a model, this was triggering a
Elementwiseerror, adding.valuesturned out to be enough.As the traces are
az.InferenceData, calculating the mean returns an array (I meant an Xarray), so added.item(0)to keep the compatibility with the formula as they exist, not sure this is the most elegant solution.Same thing for the size of the posterior distribution, you cannot just do
len(trace['var'], so I used.sizes["sample"], again not sure this is the most elegant solution, but it works.For the #152 :
sns.kdeplotfollowing seaborn depreciation warningbwtobw_methodas per seaborn depreciation warning.Simulation of field trip, as it was only doing 15 steps (index 0 is when no step have been done as far as I understood it).