#!/usr/bin/env python
# coding: utf-8

# # Python for CogSci Research

# ## Getting Python: Anaconda
# 
# https://store.continuum.io/cshop/anaconda/

# ## IPython and IPython Notebook
# 
# http://ipython.org/notebook.html

# In[1]:


3 + 5


# In[2]:


import os


# In[3]:


x = 5
y = 3
x + y


# In[4]:


import json


# In[5]:


json.dumps({"x": x, "y": y})


# ## Project structure
# 
# ```
# config.json
# analysis/
#     analyses/
#     results/
#     figures/
# bin/
# data/
#     human/
#     model
# experiment/
# lib/
# stimuli/
# ```
# 
# * `config.json`: global config file that specifies where resources are, model parameters, etc.
# * `analysis`: all scripts and results for analyzing the data
# * `bin`: helper scripts for running the experiment, running simulations, viewing stimuli, etc.
# * `data`: collected or simulated data
# * `experiment`: psiTurk experiment files
# * `lib`: importable python files that are used in analysis and model simulations
# * `stimuli`: one file per stimulus, usually JSON

# ## Running experiments
# 
# ### psiTurk
# 
# https://psiturk.org/
# 
# ### PsychoPy
# 
# http://www.psychopy.org/screenshots.html

# ## Running model simulations
# 
# ### NumPy

# In[6]:


import numpy as np


# In[7]:


np.array([7, 3, 8, 2])


# In[8]:


[7, 3, 8, 2]


# In[9]:


np.arange(10)


# In[10]:


np.arange(10) + 1


# In[11]:


range(10) + 1


# In[12]:


[x + 1 for x in range(10)]


# In[13]:


np.arange(10) + np.arange(1, 11)


# ### SciPy

# In[14]:


import scipy.stats


# In[15]:


rv = scipy.stats.norm(0, 1)
rv


# In[16]:


samples = rv.rvs(10000)
samples


# In[17]:


x = np.linspace(-3, 3)
y = rv.pdf(x)
y


# ### Matplotlib

# In[18]:


get_ipython().run_line_magic('matplotlib', 'inline')
import matplotlib.pyplot as plt


# In[19]:


plt.hist(samples, normed=True, color='k', bins=100)
plt.plot(x, y, 'r-', lw=3)


# ## Analyzing data
# 
# ### Pandas

# In[20]:


import pandas as pd


# In[21]:


data = pd.read_csv("experiment.csv")
data.head()


# In[22]:


data.set_index(["pid", "trial"]).sortlevel().head()


# In[23]:


rt = data.groupby(["HoleClass", "HoleWidth"])["responseTime"].median()
rt


# ### Matplotlib

# In[24]:


plt.bar(np.arange(len(rt)), rt, align='center')
plt.xticks(np.arange(len(rt)), rt.index)
plt.xlim(-0.75, len(rt) - 0.25)
fig = plt.gcf()
fig.set_figwidth(16)
fig.set_figheight(6)


# ### Seaborn

# In[25]:


import seaborn as sns


# In[26]:


plt.bar(np.arange(len(rt)), rt, align='center')
plt.xticks(np.arange(len(rt)), rt.index)
plt.xlim(-0.75, len(rt) - 0.25)
fig = plt.gcf()
fig.set_figwidth(16)
fig.set_figheight(6)


# In[27]:


sns.factorplot("HoleClass", "responseTime", "HoleWidth", data, estimator=np.median)
fig = plt.gcf()
fig.set_figwidth(8)
fig.set_figheight(6)


# ## Demo: responses vs. response times

# In[28]:


exp = data.groupby('isControl').get_group(False)
exp.head()


# In[29]:


rt_per_stim = exp.groupby(["stim", "HoleClass", "HoleWidth"])["responseTime"].median()
rt_per_stim


# In[30]:


resp_per_stim = exp.groupby(["stim", "HoleClass", "HoleWidth"])["response"].mean()
resp_per_stim


# In[31]:


plt.plot(resp_per_stim, rt_per_stim, 'o')
plt.xlabel("Mean response to 'will it go in?'")
plt.ylabel("Median response time")
plt.title("Responses vs. response time")


# ## Demo: R magic

# In[32]:


import rpy2


# In[33]:


get_ipython().run_line_magic('load_ext', 'rpy2.ipython')


# In[34]:


get_ipython().run_cell_magic('R', '-i exp', '\nlibrary(lme4)\n\n# check for an interaction between the hole class and hole width\nfull.lmer = lmer(responseTime ~ HoleClass*HoleWidth + (1|stim), exp)\nnoint.lmer = lmer(responseTime ~ HoleClass + HoleWidth + (1|stim), exp)\nanova(noint.lmer, full.lmer)\n')


# ## Demo: Interactive Widgets

# In[35]:


from IPython.html.widgets import interact


# In[36]:


@interact
def plot_histogram(num_samples=(0, 10000), bins=(10, 100), normed=True):
    rv = scipy.stats.norm(0, 1)
    samples = rv.rvs(num_samples)
    x = np.linspace(-3, 3)
    y = rv.pdf(x)
    plt.hist(samples, normed=normed, color='k', bins=bins)
    plt.plot(x, y, 'r-', lw=3)


# ## Other tools

# * scikit-learn
# * sympy
# * statsmodels