Search icon CANCEL
Subscription
0
Cart icon
Your Cart (0 item)
Close icon
You have no products in your basket yet
Arrow left icon
Explore Products
Best Sellers
New Releases
Books
Videos
Audiobooks
Learning Hub
Conferences
Free Learning
Arrow right icon
Arrow up icon
GO TO TOP
Bayesian Analysis with Python

You're reading from   Bayesian Analysis with Python Introduction to statistical modeling and probabilistic programming using PyMC3 and ArviZ

Arrow left icon
Product type Paperback
Published in Dec 2018
Publisher Packt
ISBN-13 9781789341652
Length 356 pages
Edition 2nd Edition
Languages
Arrow right icon
Author (1):
Arrow left icon
Osvaldo Martin Osvaldo Martin
Author Profile Icon Osvaldo Martin
Osvaldo Martin
Arrow right icon
View More author details
Toc

Table of Contents (11) Chapters Close

Preface 1. Thinking Probabilistically 2. Programming Probabilistically FREE CHAPTER 3. Modeling with Linear Regression 4. Generalizing Linear Models 5. Model Comparison 6. Mixture Models 7. Gaussian Processes 8. Inference Engines 9. Where To Go Next?
10. Other Books You May Enjoy

Single-parameter inference

In the last two sections, we learned several important concepts, but two of them are essentially the core of Bayesian statistics, so let's restate them in a single sentence.

Probabilities are used to measure the uncertainty we have about parameters, and Bayes' theorem is the mechanism to correctly update those probabilities in light of new data, hopefully reducing our uncertainty.

Now that we know what Bayesian statistics is, let's learn how to do Bayesian statistics with a simple example. We are going to begin inferring a single, unknown parameter.

The coin-flipping problem

The coin-flipping problem, or the beta-binomial model if you want to sound fancy at parties, is a classical problem in statistics and goes like this: we toss a coin a number of times and record how many heads and tails we get. Based on this data, we try to answer questions such as, is the coin fair? Or, more generally, how biased is the coin? While this problem may sound dull, we should not underestimate it. The coin-flipping problem is a great example to learn the basics of Bayesian statistics because it is a simple model that we can solve and compute with ease. Besides, many real problems consist of binary, mutually-exclusive outcomes such as 0 or 1, positive or negative, odds or evens, spam or ham, hotdog or not hotdog, cat or dog, safe or unsafe, and healthy or unhealthy. Thus, even when we are talking about coins, this model applies to any of those problems.

In order to estimate the bias of a coin, and in general to answer any questions in a Bayesian setting, we will need data and a probabilistic model. For this example, we will assume that we have already tossed a coin a number of times and we have a record of the number of observed heads, so the data-gathering part is already done. Getting the model will take a little bit more effort. Since this is our first model, we will explicitly write Bayes' theorem and do all the necessary math (don't be afraid, I promise it will be painless) and we will proceed very slowly. From Chapter 2, Programming Probabilistically, onward, we will use PyMC3 and our computer to do the math for us.

The general model

The first thing we will do is generalize the concept of bias. We will say that a coin with a bias of 1 will always land heads, one with a bias of 0 will always land tails, and one with a bias of 0.5 will land half of the time heads and half of the time tails. To represent the bias, we will use the parameter, and to represent the total number of heads for a number of tosses, we will use the variable. According to Bayes' theorem (equation 1.4), we have to specify the prior, , and likelihood, , we will use. Let's start with the likelihood.

Choosing the likelihood

Let's assume that only two outcomes are possible—heads or tails—and let's also assume that a coin toss does not affect other tosses, that is, we are assuming coin tosses are independent of each other. We will further assume all coin tosses come from the same distribution. Thus the random variable coin toss is an example of an iid variable. I hope you agree these are very reasonable assumptions to make for our problem. Given these assumptions a good candidate for the likelihood is the binomial distribution:

This is a discrete distribution returning the probability of getting heads (or in general, successes) out of coin tosses (or in general, trials or experiments) given a fixed value of :

n_params = [1, 2, 4]  # Number of trials
p_params = [0.25, 0.5, 0.75] # Probability of success

x = np.arange(0, max(n_params)+1)
f,ax = plt.subplots(len(n_params), len(p_params), sharex=True,
sharey=True,
figsize=(8, 7), constrained_layout=True)

for i in range(len(n_params)):
for j in range(len(p_params)):
n = n_params[i]
p = p_params[j]

y = stats.binom(n=n, p=p).pmf(x)

ax[i,j].vlines(x, 0, y, colors='C0', lw=5)
ax[i,j].set_ylim(0, 1)
ax[i,j].plot(0, 0, label="N = {:3.2f}\nθ =
{:3.2f}".format(n,p), alpha=0)
ax[i,j].legend()

ax[2,1].set_xlabel('y')
ax[1,0].set_ylabel('p(y | θ, N)')
ax[0,0].set_xticks(x)
Figure 1.3

The preceding figure shows nine binomial distributions; each subplot has its own legend indicating the values of the parameters. Notice that for this plot I did not omit the values on the y axis. I did this so you can check for yourself that if you sum the high of all bars you will get 1, that is, for discrete distributions, the height of the bars represents actual probabilities.

The binomial distribution is a reasonable choice for the likelihood. We can see that indicates how likely it is to obtain a head when tossing a coin (this is easier to see when , but is valid for any value of )—just compare the value of with the height of the bar for (heads).

OK, if we know the value of , the binomial distribution will tell us the expected distribution of heads. The only problem is that we do not know ! But do not despair; in Bayesian statistics, every time we do not know the value of a parameter, we put a prior on it, so let's move on and choose a prior for .

Choosing the prior

As a prior, we will use a beta distribution, which is a very common distribution in Bayesian statistics and looks as follows:

If we look carefully, we will see that the beta distribution looks similar to the binomial except for the first term, the one with all those . The first term is a normalizing constant that ensures the distribution integrates to 1, and is the Greek uppercase gamma letter and represents what is known as gamma function. We can see from the preceding formula that the beta distribution has two parameters, and . Using the following code, we will explore our third distribution so far:

params = [0.5, 1, 2, 3]
x = np.linspace(0, 1, 100)
f, ax = plt.subplots(len(params), len(params), sharex=True,
sharey=True,
figsize=(8, 7), constrained_layout=True)
for i in range(4):
for j in range(4):
a = params[i]
b = params[j]
y = stats.beta(a, b).pdf(x)
ax[i,j].plot(x, y)
ax[i,j].plot(0, 0, label="α = {:2.1f}\nβ = {:2.1f}".format(a,
b), alpha=0)
ax[i,j].legend()
ax[1,0].set_yticks([])
ax[1,0].set_xticks([0, 0.5, 1])
f.text(0.5, 0.05, 'θ', ha='center')
f.text(0.07, 0.5, 'p(θ)', va='center', rotation=0)
Figure 1.4

I really like the beta distribution and all the shapes we can get from it, but why are we using it for our model? There are many reasons to use a beta distribution for this and other problems. One of them is that the beta distribution is restricted to be between 0 and 1, in the same way our parameter is. In general, we use the beta distribution when we want to model proportions of a binomial variable. Another reason is for its versatility. As we can see in the preceding figure, the distribution adopts several shapes (all restricted to the [0, 1] interval), including a uniform distribution, Gaussian-like distributions, and U-like distributions. As a third reason, the beta distribution is the conjugate prior of the binomial distribution (which we are using as the likelihood). A conjugate prior of a likelihood is a prior that, when used in combination with a given likelihood, returns a posterior with the same functional form as the prior. Untwisting the tongue, every time we use a beta distribution as the prior and a binomial distribution as the likelihood, we will get a beta as the posterior distribution. There are other pairs of conjugate priors; for example, the Normal distribution is the conjugate prior of itself. For many years, Bayesian analysis was restricted to the use of conjugate priors. Conjugacy ensures mathematical tractability of the posterior, which is important given that a common problem in Bayesian statistics is ending up with a posterior we cannot solve analytically. This was a deal breaker before the development of suitable computational methods to solve probabilistic methods. From Chapter 2, Programming Probabilistically onwards, we will learn how to use modern computational methods to solve Bayesian problems, whether we choose conjugate priors or not.

Getting the posterior

Let's remember that Bayes' theorem (equation 1.4) says the posterior is proportional to the likelihood times the prior. So, for our problem, we have to multiply the binomial and the beta distributions:

We can simplify this expression. For our practical concerns, we can drop all the terms that do not depend on and our results will still be valid. Accordingly, we can write:

Reordering it, we get:

If we pay attention, we will see that this expression has the same functional form of a beta distribution (except for the normalization term) with and . In fact, the posterior distribution for our problem is the beta distribution:

Computing and plotting the posterior

Now we will use Python to compute and plot the posterior distribution based on the analytical expression we have just derived. In the following code, you will see there is actually one line that computes the results while the others are there just to get a nice plot:

plt.figure(figsize=(10, 8))

n_trials = [0, 1, 2, 3, 4, 8, 16, 32, 50, 150]
data = [0, 1, 1, 1, 1, 4, 6, 9, 13, 48]
theta_real = 0.35

beta_params = [(1, 1), (20, 20), (1, 4)]
dist = stats.beta
x = np.linspace(0, 1, 200)

for idx, N in enumerate(n_trials):
if idx == 0:
plt.subplot(4, 3, 2)
plt.xlabel('θ')
else:
plt.subplot(4, 3, idx+3)
plt.xticks([])
y = data[idx]
for (a_prior, b_prior) in beta_params:
p_theta_given_y = dist.pdf(x, a_prior + y, b_prior + N - y)
plt.fill_between(x, 0, p_theta_given_y, alpha=0.7)

plt.axvline(theta_real, ymax=0.3, color='k')
plt.plot(0, 0, label=f'{N:4d} trials\n{y:4d} heads', alpha=0)
plt.xlim(0, 1)
plt.ylim(0, 12)
plt.legend()
plt.yticks([])
plt.tight_layout()
Figure 1.5

On the first subplot of Figure 1.5, we have zero trials, thus the three curves represent our priors:

  • The uniform (blue) prior. This represent all the possible values for the bias being equally probable a priori.
  • The Gaussian-like (orange) prior is centered and concentrated around 0.5, so this prior is compatible with information indicating that the coin has more or less about the same chance of landing heads or tails. We could also say this prior is compatible with the belief that most coins are fair. While belief is commonly used in Bayesian discussions, we think is better to talk about models and parameters that are informed by data.
  • The skewed (green) prior puts the most weight on a tail-biased outcome.

The rest of the subplots show posterior distributions for successive trials. The number of trials (or coin tosses) and the number of heads are indicated in each subplot's legend. There is also a black vertical line at 0.35 representing the true value for . Of course, in real problems, we do not know this value, and it is here just for pedagogical reasons. This figure can teach us a lot about Bayesian analysis, so grab your coffee, tea, or favorite drink and let's take a moment to understand it:

  • The result of a Bayesian analysis is a posterior distribution, not a single value but a distribution of plausible values given the data and our model.
  • The most probable value is given by the mode of the posterior (the peak of the distribution).
  • The spread of the posterior is proportional to the uncertainty about the value of a parameter; the more spread out the distribution, the less certain we are.
  • Intuitively, we are more confident in a result when we have observed more data supporting that result. Thus, even when numerically , seeing four heads out of eight trials gives us more confidence that the bias is 0.5 than observing one head out of two trials. This intuition is reflected in the posterior, as you can check for yourself if you pay attention to the (blue) posterior in the third and sixth subplots; while the mode is the same, the spread (uncertainty) is larger in the third subplot than in the sixth subplot.
  • Given a sufficiently large amount of data, two or more Bayesian models with different priors will tend to converge to the same result. In the limit of infinite data, no matter which prior we use, all of them will provide the same posterior. Remember that infinite is a limit and not a number, so from a practical point of view, we could get practically indistinguishably posteriors for a finite and rather small number of data points.
  • How fast posteriors converge to the same distribution depends on the data and the model. In the preceding figure, we can see that the posteriors coming from the blue prior (uniform) and green prior (biased towards tails) converge faster to almost the same distribution, while it takes longer for the orange posterior (the one coming from the concentrated prior). In fact, even after 150 trials, it is somehow easy to recognize the orange posterior as a different distribution from the two others.
  • Something not obvious from the figure is that we will get the same result if we update the posterior sequentially than if we do it all at once. We can compute the posterior 150 times, each time adding one more observation and using the obtained posterior as the new prior, or we can just compute one posterior for the 150 tosses at once. The result will be exactly the same. This feature not only makes perfect sense, it also leads to a natural way of updating our estimations when we get new data, a situation common in many data-analysis problems.

The influence of the prior and how to choose one

From the preceding example, it is clear that priors can influence inferences. This is totally fine, priors are supposed to do this. Newcomers to Bayesian analysis (as well as detractors of this paradigm) are generally a little nervous about how to choose priors, because they do not want the prior to act as a censor that does not let the data speak for itself! That's OK, but we have to remember that data does not really speak; at best, data murmurs. Data only makes sense in the context of our models, including mathematical and mental models. There are plenty of examples in the history of science where the same data leads people to think differently about the same topics, and this can happen even if you base your opinions on formal models.

Some people like the idea of using non-informative priors (also known as flat, vague, or diffuse priors); these priors have the least possible amount of impact on the analysis. While it is possible to use them, in general, we can do better. Throughout this book, we will follow the recommendations of Gelman, McElreath, Kruschke, and many others, and we will prefer weakly-informative priors. For many problems, we often know something about the values a parameter can take, we may know that a parameter is restricted to being positive, or we may know the approximate range it can take, or whether we expect the value to be close to zero or below/above some value. In such cases, we can use priors to put some weak information in our models without being afraid of being too pushy. Because these priors work to keep the posterior distribution within certain reasonable bounds, they are also known as regularizing priors. Using informative priors is also a valid option if we have good-quality information to define those priors. Informative priors are very strong priors that convey a lot of information. Depending on your problem, it could be easy or not to find this type of prior. For example, in my field of work (structural bioinformatics), people have been using, in Bayesian and non-Bayesian ways, all the prior information they could get to study and especially predict the structure of proteins. This is reasonable because we have been collecting data from thousands of carefully-designed experiments for decades and hence we have a great amount of trustworthy prior information at our disposal. Not using it would be absurd! So, the take-home message is this: if you have reliable prior information, there is no reason to discard that information, including the nonsensical argument that being objective means throwing away valuable information. Imagine if every time an automotive engineer had to design a new car, they had to start from scratch and reinvent the combustion engine, the wheel, and for that matter, the whole concept of a car. That's not the way things should work.

Knowing we can classify priors into categories according to their relative strength does not make us less nervous about choosing from them. Maybe it would be better to not have priors at all—that would make modeling easier, right? Well, not necessarily. Priors can make models behave better, have better generalization properties, and can help us convey useful information. Also, every model, Bayesian or not, has some kind of prior in one way or another, even if the prior is not set explicitly. In fact, many result from frequentist statistics, and can be seen as special cases of a Bayesian model under certain circumstances, such as flat priors. One common frequentist method to estimate parameters is known as maximum likelihood; this methods avoids setting a prior and works just by finding the value of that maximizes the likelihood. This value is usually notated by adding a little hat on top of the symbol of the parameter we are estimating, such as or sometimes (or even both). is a point estimate (a number) and not a distribution. For the coin-flipping problem we can compute this analytically:

If you go back to Figure 1.5, you will be able to check for yourself that the mode of the blue posterior (the one corresponding to the uniform/flat prior) agrees with the values of , computed for each subplot. So, at least for this example, we can see that even when the maximum likelihood method does not explicitly invoke any prior, it can be considered a special case of a Bayesian model, one with a uniform prior.

We cannot really avoid priors, but if we include them in our analysis, we will get several benefits, including a distribution of plausible values and not only the most probable one. Another advantage of being explicit about priors is that we get more transparent models, meaning they're easier to criticize, debug (in a broad sense of the word), and hopefully improve. Building models is an iterative process; sometimes the iteration takes a few minutes, sometimes it could take years. Sometimes it will only involve you, and sometimes it will involve people you do not even know. Reproducibility matters and transparent assumptions in a model contribute to it. Besides, we are free to use more than one prior (or likelihood) for a given analysis if we are not sure about any special one; exploring the effect of different priors can also bring valuable information to the table. Part of the modeling process is about questioning assumptions, and priors (and likelihood) are just that. Different assumptions will lead to different models and probably different results. By using data and our domain knowledge of the problem, we will be able to compare models and, if necessary, decide on a winner. Chapter 5, Model Comparison, will be devoted to this issue. Since priors have a central role in Bayesian statistics, we will keep discussing them as we face new problems. So if you have doubts and feel a little bit confused about this discussion, just keep calm and don't worry, people have been confused for decades and the discussion is still going on.

lock icon The rest of the chapter is locked
Register for a free Packt account to unlock a world of extra content!
A free Packt account unlocks extra newsletters, articles, discounted offers, and much more. Start advancing your knowledge today.
Unlock this book and the full library FREE for 7 days
Get unlimited access to 7000+ expert-authored eBooks and videos courses covering every tech area you can think of
Renews at $19.99/month. Cancel anytime
Banner background image