3.8 The Bernoulli Distribution

Loading...

Del curso dictado por University of Pennsylvania

Fundamentals of Quantitative Modeling

3623 calificaciones

University of Pennsylvania

Fundamentals of Quantitative Modeling

3623 calificaciones

Curso 1 de 5 en SpecializationBusiness and Financial Modeling

How can you put data to work for you? Specifically, how can numbers in a spreadsheet tell us about present and past business activities, and how can we use them to forecast the future? The answer is in building quantitative models, and this course is designed to help you understand the fundamentals of this critical, foundational, business skill. Through a series of short lectures, demonstrations, and assignments, you’ll learn the key ideas and process of quantitative modeling so that you can begin to create your own models for your own business or enterprise. By the end of this course, you will have seen a variety of practical commonly used quantitative models as well as the building blocks that will allow you to start structuring your own models. These building blocks will be put to use in the other courses in this Specialization.

De la lección

Module 3: Probabilistic Models

This module explains probabilistic models, which are ways of capturing risk in process. You’ll need to use probabilistic models when you don’t know all of your inputs. You’ll examine how probabilistic models incorporate uncertainty, and how that uncertainty continues through to the outputs of the model. You’ll also discover how propagating uncertainty allows you to determine a range of values for forecasting. You’ll learn the most-widely used models for risk, including regression models, tree-based models, Monte Carlo simulations, and Markov chains, as well as the building blocks of these probabilistic models, such as random variables, probability distributions, Bernoulli random variables, binomial random variables, the empirical rule, and perhaps the most important of all of the statistical distributions, the normal distribution, characterized by mean and standard deviation. By the end of this module, you’ll be able to define a probabilistic model, identify and understand the most commonly used probabilistic models, know the components of those models, and determine the most useful probabilistic models for capturing and exploring risk in your own business.

3.1 Introduction to Probabilistic Models10:53

3.2 Examples of Probabilistic Models2:08

3.3 Regression Models4:12

3.4 Probability Trees5:00

3.5 Monte Carlo Simulations6:19

3.6 Markov Chain Models6:16

3.7 Building Blocks of Probability Models9:13

3.8 The Bernoulli Distribution7:48

3.9 The Binomial Distribution16:54

3.10 The Normal Distribution5:10

3.11 The Empirical Rule7:21

3.12 Summary2:10

Conoce a los instructores

Richard Waterman
Professor of Statistics
Statistics-Wharton School

Now it's time to look at some very specific probability distributions.

And the ones that I've chosen here are probably the most frequently used

probability distributions.

And it is the case that there are hundreds of different probability distributions

out there.

And I've chosen three to present, but they are very very foundational.

And if you were setting up a Monte Carlo simulation it's quite likely that a lot of

the inputs would have

one of the following distributions that I'm going to talk about now.

So here's the most elemental of all the probability distributions,

it's called the Bernoulli distribution.

And what the Bernoulli distribution models for

you is a random variable that can only take on one of two values.

So, the easiest way to think about this is tossing a coin,

it can come up as a head or a tail.

But, more generally, we all represent the outcomes as either a one or a zero.

So, you might equate a head to a one and a tail to a zero, and so

we've got an event that can only take on one of two outcomes and

we call the probability that x equal to one, we give that the letter p in general.

And because something has to happen that means the probability x equal to zero is

one minus p and so that's a very sort of elemental or

a comic random variable from which we'd be able to build up lots of more

Interesting and complicated probability distributions.

So often times when we talk about these Bernoulli random variables we view them as

an outcome of an experiment and

that outcome can either be a success or a failure.

And in a business setting that might be whether an individual buys your product or

doesn't buy your product.

And as an example of a Benoulli random variable.

Now, once we got this probability distribution written down,

we're able to calculate the summaries.

So I'm going to introduce the summaries now.

The first one is mu, the mean.

Another way of writing that is the expected value of X,

the average or the expected value.

So you can see E of X is another way of denoting the mean,

the expected value of X and the way that you get the expected value of one of these

discrete random variables Is by saying, what values can it take, and

what are the probability that it takes on those specific value and then add them up.

And so this Bernoulli can take on the value one with probability p, or

it can take on the value zero with probability one minus p.

If you do that algebra,

you'll find that the mean of this Bernoulli random variable is p.

Now the definition of the variance sigma squared

is the expected value of X minus mu squared.

So you can see through this formula why the variance is able to

pick up the spread.

Because if you look inside the parenthesis,

you'll see mu there, that's the center of the distribution.

And X- mu is telling you how far a particular

value of the random variable is away from the center.

So how far are you from the center and the expected value of that quantity squared is

sort of how far on average are you from the center, but on a squared level.

So using a square distance.

Now if you work that quantity out the expectation calculation is the same idea.

You figure out the values that X- mu can take and

then you multiply those values by the probabilities and so

if we work that out, we'll see that X- mu, that would be 1- p,

X- mu squared times p is the first term because if X = 1,

then you just plug 1 in and you get 1- mu squared, and

what's the probability that X = 1?

That's p, so that's the first term, 1- p squared times p, and

then the other possible outcome is for X to equal 0.

And if you put 0 in and you do 0-mu, but remember mu,

the mean is just p from the original, the previous calculation.

You get 0-p all squared, times the probability that you got a 0,

which is 1-p.

And so you can see the two probabilities there, p and 1- p.

And you can see the two components.

What's X- mu squared?

It's either 1 p all squared or 0- p all squared.

You take that.

You do your algebra, and you end up with p times 1- p.

So that is the variance of a Bernoulli.

And if we want to get the standard deviation

of this particular random variable, then that's easy.

You just take the square root of the variance.

And so those are the calculations associated with finding a mean and

a variance or standard deviation.

If you're wondering why we have these two different measures of spread, the variance

and the standard deviation, it's really about the units associated with them.

So let's say X was measured in yen, for the sake of argument,

then the variance would be measured in yen squared, so

it's a little bit hard to get your head around the idea of a squared yen.

But if you square root that and

you look at the units of measurement, if the variance is measured in yen squared,

then the standard deviation would be measured in yen, so it's back on the scale

of the data, so it's easier to interpret the standard deviation.

So, those are our calculations for the key summaries, and

just to do a particular example, what if we have p = 0.5?

So that's a very special instance of the Bernoulli,

where each outcome is equally likely.

That's tossing a fair coin.

Then you'll find that for p = 0.5,

the mean is just equal to 0.5 itself, remember the mean is equal to p.

The variance is p times 1-p, a half times a half is a quarter,

and then the standard deviation if you square root the quarter,

you are going to get a half, and so, those are the key summaries.

And, this process of creating the summaries is very useful as I say

at the end of having done something like a Monte Carlo simulation to say,

here's why I think the answer is going to be mu and then we can use the standard

deviation to give us a sense of the variability about the mean.

So here's a example of where this Bernoulli distribution could come

in useful.

And I'm going back to one of the initial examples of this class,

where we're talking about drug development and creating a probabilistic model

would potentially involve these Bernoulli random variables.

So we might come up with a scenario where we think the outcome of whether or

not a drug is approved by some regulatory body is a Bernoulli random variable.

It's either approved yes or it's not approved, no.

We can write those as the values 1 and 0 so we're back to this Bernoulli setup.

And then our experts might have estimated that the probability that we're a yes,

that we get approved, is 0.65 and the probability we don't get approved is 0.35.

Notice those two add up to one.

Now if we believe that given the drug is approved,

the projected revenue is $500 million from it, zero otherwise because

if it's not approved we can't sell it, so we'll get no revenue at all.

We could work out the expected revenue.

And the calculation of these expectations is as before you take the different

realizations of the random variable and the revenues under each realization.

More applied through the probabilities of those events.

So if we get approved,

which is probability 0.65, we get 500 million revenue.

And if we don't get approved we get zero revenue, and

that happens with probability 0.35.

We add those up and we get an expected revenue of $325 million.

And so this Bernoulli could potentially be used as a building block for

a Monte Carlo simulation to get a sense of,

if we've got ten of these drugs what's the expected revenue going to be?

Explora nuestro catálogo

Inscríbete de manera gratuita y obtén recomendaciones personalizadas, actualizaciones y ofertas.

Coursera brinda acceso universal a la mejor educación del mundo, al asociarse con las mejores universidades y organizaciones, para ofrecer cursos en línea.

© 2019 Coursera Inc. Todos los derechos reservados.

Descargar en la App Store

Consíguelo en Google Play