New argument for plotting true values in pairplots #1140
Conversation
|
Hi, do we have some other possible name that I get the meaning of the |
| true_values_copy = {} | ||
| label = [] | ||
| for variable in list(true_values.keys()): | ||
| if " " in variable: |
There was a problem hiding this comment.
why there is this step?
There was a problem hiding this comment.
Hi! I was trying to figure out a way so that the user can specify variable names in an easy manner. For example, simply writing "theta Choate" instead of "theta\nChoate"
CHANGELOG.md
Outdated
| * Add out-of-sample groups (`predictions` and `predictions_constant_data`) and `constant_data` group to pyro translation #1090 | ||
| * Add `num_chains` and `pred_dims` arguments to io_pyro #1090 | ||
| * Integrate jointplot into pairplot, add point-estimate and overlay of plot kinds (#1079) | ||
| * Allow xarray.Dataarray input for plots.(#1120) | ||
| * Revamped the `hpd` function to make it work with mutidimensional arrays, InferenceData and xarray objects (#1117) | ||
| * Skip test for optional/extra dependencies when not installed (#1113) | ||
| * Add option to display rank plots instead of trace (#1134) | ||
| * Integrate jointplot into pairplot, add point-estimate and overlay of plot kinds #1079 |
There was a problem hiding this comment.
This line is repeated 3 times
| @@ -140,6 +175,12 @@ def plot_pair( | |||
|
|
|||
| ax.scatter(pe_x, pe_y, marker="s", s=figsize[0] + 50, **point_estimate_kwargs, zorder=4) | |||
|
|
|||
| if true_values: | |||
| ax.scatter( | |||
There was a problem hiding this comment.
Why scatter instead of plot?
| @@ -88,6 +120,9 @@ def plot_pair( | |||
| for val, ax_, rotate in ((x, ax_hist_x, False), (y, ax_hist_y, True)): | |||
| plot_dist(val, textsize=xt_labelsize, rotated=rotate, ax=ax_, **marginal_kwargs) | |||
|
|
|||
| ax_hist_x.set_xlim(ax.get_xlim()) | |||
There was a problem hiding this comment.
I think this was removed recently. It was an error deleting it or adding it again is a mistake?
There was a problem hiding this comment.
I think I deleted it. ax_hist_x is created with sharex=ax, therefore axis limits and ticks will be the same, there is no need to manually set both limits after plotting.
|
I agree the name should me more general. May be ref_value, reference_value, ref_point? |
There was a problem hiding this comment.
plot_posterior uses ref_val, we could stick with that and try to mimic its structure (i.e. allows several ref values per variable/coord) as much as possible adapting the behaviour?
When I first saw the issue, my first take to approach is was to call 2 times plot_pair using the same axis:
idata = az.load_arviz_data("centered_eight")
ax = az.plot_pair(idata, group="posterior_predictive")
az.plot_pair(idata, group="observed_data", ax=ax, scatter_kwargs=dict(color="red"));
which can also be used to plot crossed lines instead of a single marker:
which then got me thinking, would it be useful to have some plot_ref_vallike functions? This is probably not the best place to share, but I'll create an issue if there is quorum.
| @@ -88,6 +120,9 @@ def plot_pair( | |||
| for val, ax_, rotate in ((x, ax_hist_x, False), (y, ax_hist_y, True)): | |||
| plot_dist(val, textsize=xt_labelsize, rotated=rotate, ax=ax_, **marginal_kwargs) | |||
|
|
|||
| ax_hist_x.set_xlim(ax.get_xlim()) | |||
There was a problem hiding this comment.
I think I deleted it. ax_hist_x is created with sharex=ax, therefore axis limits and ticks will be the same, there is no need to manually set both limits after plotting.
Codecov Report
@@ Coverage Diff @@
## master #1140 +/- ##
==========================================
+ Coverage 92.81% 92.98% +0.17%
==========================================
Files 93 93
Lines 9181 9229 +48
==========================================
+ Hits 8521 8582 +61
+ Misses 660 647 -13
Continue to review full report at Codecov.
|
…es (#1079) * add jointplot features into pairplot * add scatter_kde kind for pairplot * add point_estimate arguments * bokeh backend * fix None argument for color in kdeplot bokeh backend * run black, pylint and pytest * remove scatter_kde kind among several other changes * minor changes * run pytest * add plot width and height to backend_kwargs fix pylint issues fix hover feature fix hover feature minor fixes * update docstring run pylint * update changelog
remove whitespace
change argument name to reference_values
agustinaarroyuelo
changed the title
Description
This PR adds the argument
true_valuesinplot_pair. It allows for scattering the true values of a variable as suggested in #1131.Checklist