matplotlib-users

[Matplotlib-users] for Posterior predictive data distributions

[Juan Wu <wuj...@gm...>]

Hi, Experts,

My colleagues and I have a question, how we can make a plot via python like
below. According to a guy's original paper, "Each panel shows the
normalized histograms of the observed data (bar plots) and the model
prediction (black lines) ".

I believe that people can make it with Matplotlib. Any code suggestion
(with simple example data) would be much appreciated.

(I am more comfortable with Matlab, but now the python code is preferred).

J


[image: Inline image 3]

Re: [Matplotlib-users] for Posterior predictive data distributions

[Paul H. <pmh...@gm...>]

(apologies if the list receives this twice)

On Fri, Jun 5, 2015 at 9:14 AM, Juan Wu <wuj...@gm...> wrote:
>
>> Hi, Experts,
>>
>> My colleagues and I have a question, how we can make a plot via python
>> like below. According to a guy's original paper, "Each panel shows the
>> normalized histograms of the observed data (bar plots) and the model
>> prediction (black lines) ".
>>
>> I believe that people can make it with Matplotlib. Any code suggestion
>> (with simple example data) would be much appreciated.
>>
>> (I am more comfortable with Matlab, but now the python code is preferred).
>>
>> J
>>
>

Juan,

It is, of course, very difficult to give any concrete advice without
knowing how your data are stored.

In any case, seaborn builds on matplotlib to provide some very advanced
visualization with a very concise API.

I recommend you look into the seaborn.distplot function and
seaborn.FacetGrid class.
http://web.stanford.edu/~mwaskom/software/seaborn/

-Paul

Re: [Matplotlib-users] for Posterior predictive data distributions

[Jan H. <jan...@gm...>]

I put the data into a list of lists of numpy arrays. The following script
generated a plot similar to what Juan attached:

import matplotlib.pyplot as plt
import numpy as np

def model(t, ii, jj):
    """
    Returns some numbers according to the independent variable, t, and
    parameters of the model, ii and jj.
    """
    return (jj*6+ii)*np.ones_like(t)

### generate data (i.e. model + some random values)
t = np.linspace(-2.5, 2.5, 11)
data = [
    [model(t, ii, jj) + np.random.rand(len(t)) - .5
     for ii in range(6)] # different sessions
    for jj in range(2)] # accuracy/speed

### do the plotting
colors = ['b', 'r']
titles = ['Accuracy emphasis', 'Speed emphasis']
fig, big_axes = plt.subplots(figsize=(16, 6), nrows=2, ncols=1, sharey=True)

for row, big_ax in enumerate(big_axes):
    big_ax.set_ylabel('RT distributions')
    big_ax.set_xlabel('Response time [s]')
    big_ax.set_title(titles[row])
    big_ax.tick_params(labelcolor=(1.,1.,1.,0.), top='off', bottom='off',
left='off', right='off')
    big_ax._frameon = False

for jj, emph_data in enumerate(data):
    for ii, iidata in enumerate(emph_data):
        ax = fig.add_subplot(len(data), len(emph_data),
jj*len(emph_data)+ii+1)
        # plot bars:
        ax.bar(t, iidata,
               color=colors[jj],
               width=(max(t)-min(t))/(len(t)-1), # for non-overlapping bars
               linewidth=0) # no outlines
        # plot models:
        ax.plot(t, model(t, ii, jj), 'k', linewidth=2)

        # annotate axes etc.
        ax.set_xticks((-2.5, 0., 2.5))
        ax.set_ylim((-.1, 12.8))
        ax.annotate('Session {}'.format(ii+1), xy=(-2.4, 11.8))

plt.tight_layout()
plt.show()

How to add the centered titles for each row is described here :
http://stackoverflow.com/questions/27426668/row-titles-for-matplotlib-subplot

Jan

On 8 June 2015 at 03:27, Paul Hobson <pmh...@gm...> wrote:

> (apologies if the list receives this twice)
>
> On Fri, Jun 5, 2015 at 9:14 AM, Juan Wu <wuj...@gm...> wrote:
>>
>>> Hi, Experts,
>>>
>>> My colleagues and I have a question, how we can make a plot via python
>>> like below. According to a guy's original paper, "Each panel shows the
>>> normalized histograms of the observed data (bar plots) and the model
>>> prediction (black lines) ".
>>>
>>> I believe that people can make it with Matplotlib. Any code suggestion
>>> (with simple example data) would be much appreciated.
>>>
>>> (I am more comfortable with Matlab, but now the python code is
>>> preferred).
>>>
>>> J
>>>
>>
>
> Juan,
>
> It is, of course, very difficult to give any concrete advice without
> knowing how your data are stored.
>
> In any case, seaborn builds on matplotlib to provide some very advanced
> visualization with a very concise API.
>
> I recommend you look into the seaborn.distplot function and
> seaborn.FacetGrid class.
> http://web.stanford.edu/~mwaskom/software/seaborn/
>
> -Paul
>
>
> ------------------------------------------------------------------------------
>
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>
>

Re: [Matplotlib-users] for Posterior predictive data distributions

[Thomas C. <tca...@gm...>]

Juan,

If you join the mailing list you will be able to post without moderation.

Those look like a combination of `ax.hist` or `ax.bar` + `ax.plot(x, y,
lw=3, c='k')` + `fig, ax_list = plt.subplots(1, 6)` + turning off some of
the spines + `fig.text` for the shared axes labels and `fig.suptitle` for
the titles.

Tom

On Sun, Jun 7, 2015 at 3:06 PM Juan Wu <wuj...@gm...> wrote:

> Hi, Experts,
>
> My colleagues and I have a question, how we can make a plot via python
> like below. According to a guy's original paper, "Each panel shows the
> normalized histograms of the observed data (bar plots) and the model
> prediction (black lines) ".
>
> I believe that people can make it with Matplotlib. Any code suggestion
> (with simple example data) would be much appreciated.
>
> (I am more comfortable with Matlab, but now the python code is preferred).
>
> J
>
>
> [image: Inline image 3]
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>

Re: [Matplotlib-users] for Posterior predictive data distributions

[Paul H. <pmh...@gm...>]

Juan,

It is, of course, very difficult to give any concrete advice without
knowing how your data are stored.

In any case, seaborn builds on matplotlib to provide some very advanced
visualization through very concise code.

I recommend you look into the seaborn.distplot function and
seaborn,FacetGrid class.
http://web.stanford.edu/~mwaskom/software/seaborn/

-Paul

On Fri, Jun 5, 2015 at 9:14 AM, Juan Wu <wuj...@gm...> wrote:

> Hi, Experts,
>
> My colleagues and I have a question, how we can make a plot via python
> like below. According to a guy's original paper, "Each panel shows the
> normalized histograms of the observed data (bar plots) and the model
> prediction (black lines) ".
>
> I believe that people can make it with Matplotlib. Any code suggestion
> (with simple example data) would be much appreciated.
>
> (I am more comfortable with Matlab, but now the python code is preferred).
>
> J
>
>
> [image: Inline image 3]
>
>
> ------------------------------------------------------------------------------
>
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users
>
>