Clean up GLM doc

[junpenglao]

• pm.glm.glm -> pm.glm.GLM.from_formula
• pm.glm.plot_posterior_predictive -> pm.plot_posterior_predictive_glm
• remove find_MAP and used default ADVI init

[junpenglao]

@twiecki I assigned you as you were working on this quite a lot before ;-)

[twiecki]

Looks great!

Can you delete the cell:

def f(x, y):
    print(x)
    print(y)
    
f(y=1, x=2)

in GLM-robust-with-outlier-detection.ipynb

Then GLM-logistic.ipynb also seems to have some exceptions in some cells, probably only needs to be rerun.

[junpenglao]

OK should be fine now.
The only problem: in GLM-rolling-regression.ipynb the YahooDailyReader is currently broken (see pydata/pandas-datareader#333 and pydata/pandas-datareader#315). I did not rerun the notebook there.

[twiecki]

@junpenglao Did you try loading via Google? Interface is identical via pandas-datareader.

[junpenglao]

No as I never use Pandas-datareader. Trying now!

[ColCarroll]

that looks like a small file -- would it be better to download, check into examples/data, and read with pd.read_csv(pm.read_data('stock_prices.csv'))?

[twiecki]

@ColCarroll Good point.

[junpenglao]

@twiecki Using GoogleDailyReader there is no 'Adj Close' (prices.items -> Index(['Open', 'High', 'Low', 'Close', 'Volume'], dtype='object')), and the GDX data in 'Close' are NaNs.

@ColCarroll I agree.

[twiecki]

This looks great, thanks!

[junpenglao]

Also, i find that Metropolis performs surprisingly well in GLM - in many cases out performs NUTS

[twiecki]

In low dimensional models?

[junpenglao]

yeah in GLM most model only has 2-3 hierarchies.