Ch04 "Doing the Work Yourself": typo in code

[nobutoba]

I'm having fun reading your book and found a small typo (the code does not acctually appear in your GitHub repo, though).

In fact, the function open_rate_with_factor_change defined on p.123 does not work as expected. One way to fix this is to change the following code:

    num_opened = np.zeros(len(period_rng))
    for prd in range(0, len(period_rng), 2): 
        try:
            n, p = NUM_EMAILS_SENT_WEEKLY, np.random.uniform(0,
                                                             1)
            num_opened[prd: (prd + 2)] = np.random.binomial(n, p,
                                                            2)
            p = max(min(1, p * fac), 0)
        except:
            num_opened[prd] = np.random.binomial(n, p, 1)

into something like:

    num_opened = np.zeros(len(period_rng))
    n, p = NUM_EMAILS_SENT_WEEKLY, np.random.uniform(0,
                                                     1)
    for prd in range(0, len(period_rng), 2): 
        try:
            num_opened[prd: (prd + 2)] = np.random.binomial(n, p,
                                                            2)
            p = max(min(1, p * fac), 0)
        except:
            num_opened[prd] = np.random.binomial(n, p, 1)

The point here is that random sampling of the probability p should occur outside of the for loop, in order for the increasing or decreasing of email open rate acctually takes place.