def group_by_columns(data, group_by_columns, column_to_agg):
'''Takes a list of column names and a column to count and counts rows by those column name, including nulls'''
all_columns = group_by_columns + [column_to_agg]
data.loc[:, column_to_agg] = data.loc[:, column_to_agg].fillna(value=0)
data = data.loc[:, all_columns]
grouped = data.groupby(group_by_columns)
return grouped
def fit_polynomial(data, title, file, alpha=0):
"""
Fits data to linear models of polynomial degrees 1-5 to compare
:param data: water maze cipl data
:param title: title of plot (string)
:param file: filename
:param alpha: alpha value
:return: predictions, y intercept, coefficients, and errors
"""
X_train, X_test, y_train, y_test = train_test_split(
data['Trial'], data['Water Maze CIPL'])
colors = ['teal', 'yellowgreen', 'gold', 'purple', 'pink', 'brown']
lw = 2
mses = []
for degree in [1, 2, 3, 4, 5]:
model = make_pipeline(PolynomialFeatures(degree), Ridge(alpha=alpha))
model.fit(X_train.values.reshape(-1, 1), y_train)
if degree == 1:
line = model.predict(np.linspace(0, 30, 100).reshape(-1, 1))
intercept = model._final_estimator.intercept_
coef = model._final_estimator.coef_
y_plot = model.predict(np.linspace(0, 30, 100).reshape(-1, 1))
mse = mean_squared_error(y_test,
model.predict(X_test.values.reshape(-1, 1)))
mses.append(mse)
plt.plot(np.linspace(0, 30, 100),
y_plot,
color=colors[degree - 1],
linewidth=lw,
label="degree {0} test error:{1}".format(
degree, mse.round(2)))
ys = data.groupby('Trial').mean()['Water Maze CIPL']
plt.scatter(range(1, 25),
ys,
color='black',
edgecolors='black',
s=30,
marker='o',
label="Average trial performance")
plt.legend(loc=2, prop={'size': 8})
plt.title(title)
plt.xlim(0, 30)
plt.ylim(0, 60)
plt.xlabel('Trial')
plt.ylabel('CIPL')
plt.savefig('Results/Regression/Learning/' + file)
plt.show()
return line, intercept, coef, mses
def create_counts(data, group_by_columns, column_to_count):
'''Takes a list of column names and a column to count and counts rows by those column name, including nulls'''
all_columns = group_by_columns + [column_to_count]
data = data.loc[:, all_columns].fillna(value=1)
grouped = data.groupby(group_by_columns)
return grouped.count()