added data perprocessing and linear model file #33
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
| @@ -0,0 +1,717 @@ | |||
| { | |||
There was a problem hiding this comment.
Can you add some headers to the sections in this notebook like you have in the other one (Dowload data, Detrend data, etc)
Reply via ReviewNB
| @@ -0,0 +1,717 @@ | |||
| { | |||
There was a problem hiding this comment.
Instead of dropping the trends later, I think it would be better to make df_trend = df.copy() here, then save and plot them separately:
for col in log_cols: cyc, trend = hpfilter(np.log(df[col].dropna()), lamb=1600) df.loc[df[col].dropna().index, col] = cyc df_trend.loc[df[col].dropna().index, col] = trend
Reply via ReviewNB
| @@ -0,0 +1,717 @@ | |||
| { | |||
There was a problem hiding this comment.
There's a helper function in gEconpy.plotting called prepare_gridspec_figure,it sets up arrays of subplots for non-square numbers of plots, and will prevent you from having these two empty ones at the end. You can see how it is used here: https://github.com/jessegrabowski/gEconpy/blob/cd2ae75747b0aa8bc23a6b35033f3a3e4ca7bc79/gEconpy/plotting/plotting.py#L237
Reply via ReviewNB
| @@ -0,0 +1,717 @@ | |||
| { | |||
There was a problem hiding this comment.
I like the heatmaps you use in the other notebook better than just a table, so the viewer's eye can follow the colormap. Personally, blocks of numbers just wash over me like I'm on ketamine.
Reply via ReviewNB
| @@ -0,0 +1,376 @@ | |||
| { | |||
There was a problem hiding this comment.
| @@ -0,0 +1,376 @@ | |||
| { | |||
There was a problem hiding this comment.
Line #2. print('#' * 20, 'SOLVE MODEL', '#'*20)
For print like this you can do print(' SOLVE MODEL '.center(50, '#')) (the number is the total line length, including the SOLVE MODEL part
Reply via ReviewNB
| @@ -0,0 +1,376 @@ | |||
| { | |||
There was a problem hiding this comment.
Can you pass shock sizes to silence those warnings? Even if you just use the default. Since they're all the same you can just pass a covariance matrix:
mod1.compute_stationary_covariance_matrix(shock_cov_matrix=np.eye(3) * 0.01)
Reply via ReviewNB
| @@ -0,0 +1,376 @@ | |||
| { | |||
There was a problem hiding this comment.
Line #3. mask[np.triu_indices_from(mask)] = True
Maybe use k=1 in triu_indices_from so that you get the main diagonal in the plots, as in mask[np.triu_indices_from(mask, k=1)] = True ? The variance is as interesting as the covariances.
Reply via ReviewNB
| @@ -0,0 +1,376 @@ | |||
| { | |||
There was a problem hiding this comment.
Line #1. from src.utils import plot_dfs
I think it looks nicer to keep all the imports at the top of the notebook
Reply via ReviewNB
| @@ -0,0 +1,376 @@ | |||
| { | |||
There was a problem hiding this comment.
Line #7. plot_dfs(dfs, sns.lineplot, legend=[i.base_name for i in mod.shocks]);
There's a plotting function for IRFs in gEconpy.plotting.plot_irf,was it not suitable for your needs? I should tune it up if so, so you don't need to write any custom plotting functions in a notebook like this.
Reply via ReviewNB
There was a problem hiding this comment.
This is really great! I made a couple small comments. I think it would be good if you could roll all 3 files (the two notebooks and the utils) into a single file. My dream is to set up an "example gallery" like PyMC/Arviz has, so these two notebooks combined would be a great addition to that.
If you have extra time, my "stretch goal" would be a bit of commentary/analysis of the results. That would give it a bit more pedagogical value for someone who wants to do an analysis like this. Let me know if you want some help writing that up.
examples/src/utils.py
Outdated
| from itertools import chain | ||
|
|
||
|
|
||
| def plot_dfs(dfs, plot_func, dfs_cov: pd.DataFrame=None, when_crisis: pd.Series=None, legend: list=None, no_cols: int=3, figsize: tuple=(5, 2), **kwargs): |
There was a problem hiding this comment.
Is it possible to use the "built-in" plotting functions? It'd be nice to avoid a utils file for an example notebook, and as I remarked in the notebook, if those functions don't do what you need them to do, I need to fix them up
| ########################################################################### | ||
| # New Keynesian model (Galì, 2008) | ||
| ########################################################################### | ||
|
|
There was a problem hiding this comment.
I added an options flag you can use in linear models that will automatically pass linear_model = True to places like steady_state and solve_model. You can see an example here: https://github.com/jessegrabowski/gEconpy/blob/main/tests/Test%20GCNs/RBC_Linearized.gcn
|
Also you can ignore the CI failure, it has nothing to do with your PR |
I have added some data preprocessing and the Galì 2008 canonical DSGE in linear as well as FOC form.