rouseguy / intro2stats Goto Github PK

In the shuffle_experiment function, I believe the accessing of the experiment_data array used to calculate the mean is off (see highlighted portion below). Specifically, the condition is grabbing the rows labeled 0 or 1 correctly, however, after that we need to only grab the second column of each row so as to exclude the label value from the mean calculation.

Current code:

def shuffle_experiment(number_of_times):
experiment_diff_mean = np.empty([number_of_times,1])
for times in np.arange(number_of_times):
experiment_label = np.random.randint(0,2,shoe_sales.shape[0])
experiment_data = np.array([experiment_label, shoe_sales[:,1]]).T
experiment_diff_mean[times] = experiment_data[experiment_data[:,0]==1].mean()
- experiment_data[experiment_data[:,0]==0].mean()
return experiment_diff_mean

Proposed code:

def shuffle_experiment(number_of_times):
experiment_diff_mean = np.empty([number_of_times,1])
for times in np.arange(number_of_times):
experiment_label = np.random.randint(0,2,shoe_sales.shape[0])
experiment_data = np.array([experiment_label, shoe_sales[:,1]]).T
experiment_diff_mean[times] = experiment_data[experiment_data[:,0]==1][:,1].mean()
- experiment_data[experiment_data[:,0]==0][:,1].mean()
return experiment_diff_mean

The same issue exists in this block:

experiment_diff_mean = experiment_data[experiment_data[:,0]==1][:,1].mean()
- experiment_data[experiment_data[:,0]==0][:,1].mean()

In the notebook 2 "Warm-up", the following statement is wrong:

Simulation. Run the experiment 100,000 times. Find the percentage of times the experiment returned 24 or more heads. If it is more than 5%, we conclude that the coin is biased.

The statement should be

Simulation. Run the experiment 100,000 times. Find the percentage of times the experiment returned 24 or more heads. If it is less than 5%, we conclude that the coin is biased.

This error was introduced by commit 1c3c897

In your linear regression Notebook, you mentioned about scikit-learn, but as we know that does not have good statistical tests for the linear regression (residual analysis or F-test etc. like R does).

I have developed a lightweight package called MLR which mixes the simple fit and predict of the scikit-learn package with statistical inference and residual analysis.

You can refer to it or use it in your Notebooks if you like. Here are the docs.

I am running Ubuntu 14.04 64 bit on a machine with Python 2.7 in the venv. Whenever I am trying to install requirements.txt or requirements_linux.txt, I am getting the following error

Could not find a version that satisfies the requirement functools32==3.2.3.post2 (from -r requirements.txt (line 6)) (from versions: 3.2.3-1, 3.2.3-1, 3.2.3-2, 3.2.3-2)

I bypassed this error by simply writing

functools32==3.2.3-2

I think both the requirements should be updated or is just a error specific to my machine only?