Plotting issue

[ThiagoBorges81]

Hello, Todd.

I hope you are well. Please, I was wondering if you could point me towards a direction to format the plot for the anova2nested. The labels for the p-values, the alphas, and the F stats are overlapping in the plot, which makes it difficult to see and understand the effects. Also, I would like to extend the x-axis so that the number 40 also appears in the axis. Please, see a sample of what is happening:

I searched for the code that generate the plots, but I could not find it in the directories where spm1d is installed. Any assistance would be appreciated.

Kind regards

Thiago

[0todd0000]

spm1d supports just basic plotting by default so you'd need to customize things a bit...

I would like to extend the x-axis so that the number 40 also appears in the axis

Try using set_xlim like this:

import matplotlib.pyplot as plt
import spm1d

dataset = spm1d.data.uv1d.anova2nested.SPM1D_ANOVA2NESTED_3x3()
Y,A,B   = dataset.get_data()
FF      = spm1d.stats.anova2nested(Y, A, B, equal_var=True).inference(0.05)
F       = FF[0]

plt.figure()
ax = plt.axes()
F.plot( ax=ax )
ax.set_xlim(0, 40)
plt.show()

The labels for the p-values, the alphas, and the F stats are overlapping in the plot, which makes it difficult to see and understand the effects.

Either (a) use the offsets keyword argument to offset the p-value labels from their default positions (the default position is the cluster centroid). For example

# offsets in vertical direction
# --- number of 2-tuples must be the same as the number of clusters
offsets = [(0,3), (0,4), (0,5), (0,6)]  
F.plot_p_values( ax=ax, offsets=offsets )

or (b) you could create your own custom version of plot_p_values whose source code is available here.

[ThiagoBorges81]

Hello, Todd.

Thank you very much for getting back to me. Most appreciated.

The solutions you provided are pointing me to the class SPMiPlotter source code. However, the spm1d package I have installed is using the class SPMList. This class is neither in the _plot.py nor in the plot.py files.

This is the version I have installed: spm1d==0.4.28

Please, find below the code I am using:

#(1) Conduct ANOVA:
alpha        = 0.05
FF           = spm1d.stats.anova2nested(Y_vel, A_vel, B_vel, equal_var=True)
FFi          = FF.inference(0.05)
print( FFi )

#(2) Plot results:
plt.close('all')
FFi.plot(plot_threshold_label=True, plot_p_values=True)
plt.show()

And here is the output:

As I was not getting the results I expected, I threw a random argument to double-check which class the package was calling, and I got the following error:

{
    "name": "TypeError",
    "message": "SPMFList.plot() got an unexpected keyword argument 'color'",
    "stack": "---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[67], line 14
      8 #(2) Plot results:
      9 # plt.close('all')
     10 # plt.figure()
     11 # plt.axes()
     13 offsets = [(0,3),(0,4),(0,5),(0,6),(0,7)]
---> 14 fig = FFi.plot(plot_threshold_label=True, plot_p_values=True, color='g')
     16 plt.show()

TypeError: SPMFList.plot() got an unexpected keyword argument 'color'"
}

So, it is pointing to the SPMFList class instead of the SPMiPlotter.

Any help would be appreciated.

Kind regards

Thiago

[0todd0000]

Understood, thank you for clarifying. Note that FFi.plot calls _plot_F_list which is in _plot.py here. You'll see in _plot_F_list that the p-value plotting function is the same as the one above: F.plot_p_values. Please note that the FFi.plot method (like all plotting methods in spm1d) is just meant to be a convenience visualization method and is not meant to offer comprehensive figure customization options. When plotting from a list of F objects it is particularly difficult to offer separate customizations for each F plot, so if you wish to customize the default plot I suggest creating a multi-panel figure, plotting each F object in a separate panel, and then customizing that panel as desired. You can use _plot_F_list as a template if you wish but I think it might be easiest just to call something like plt.subplots and the customizing.