The Box Model Part 3 - Modelling Using a Normal Distribution

Modelling Using a Normal Distribution

In Part 2, we met the central limit theorem: if each sample is made up of a sufficiently large number of draws from a box, then the sample means (or sums) will follow an approximately normal distribution. We also saw how to check whether the number of draws is large enough — repeatedly form samples, calculate the mean (or sum) of each one, and see whether the histogram of those values looks normal.

Why do we care about that shape? Because once we know the sample sums (or means) follow a normal distribution, we can start to answer probability-based questions about them.

For example: what is the probability of flipping 60 or more heads out of 100 coins? We start by writing this as a box model. We put a '1' ticket in the box (representing a head) and a '0' ticket (representing a tail). Flipping 100 coins is the same as drawing 100 times from that box, and the number of heads is simply the sum of the 100 tickets we draw. This box model is pictured to the right.

To turn that box into a probability, we work through four steps. These four steps are exactly the four sections of the Box Model Playground below, in the order they appear:

  1. Box Parameters — describe the box: which tickets go in it, how many draws make up a sample, and whether we summarise a sample by its sum or its mean. (For our coin question: tickets 1 and 0, n = 100 draws, and the sum.)
  2. Central Limit Theorem — check that n = 100 draws really is enough for the sample sums to look normal. We do this empirically, by repeatedly drawing samples and building up a histogram of their sums.
  3. Modelling Using a Normal Distribution — pin down which normal curve to use. A normal distribution needs a mean and a standard deviation, so we use the expected value (EV) of the sample sum as the mean, and its standard error (SE) as the standard deviation (we will learn how to calculate EV and SE below). (For our coin question this gives EV = 50 and SE = 5, so we model the sample sum as \(N(50, 5^2)\).)
  4. Finding Probabilities — read the probability off that curve as an area underneath it. (For '60 or more heads', we shade the area to the right of 60, because values above 60 mean more than 60 heads. So we set the lower boundary to 60, and tick the \(\infty\) checkbox for the upper boundary — there is no upper limit on the number of heads we are counting. That area is 0.02275, so the chance of flipping 60 or more heads from 100 coins is 2.275%.)

Don't worry if some of this still feels vague — each of the four sections below explains its own step in much more depth, and updates itself automatically for whatever box you set up. The playground already starts configured for the coin question above, so the best first move is to scroll down and follow the four sections through in order.

The first suggested exercise is to recreate the scenario above, but with a slight change. What is the probability that if you flipped 100 coins, you would observe 70 or more heads?

To do this:

  • In the 'Box Parameters' mini-section, set the tickets and number of draws to match the box model to the right. Ensure that we take the sum for the samples (you could also take the mean, if you prefer, but you will need to adjust later steps).
  • Verify that n = 100 is a sufficient number of draws for the central limit theorem to apply for this box model.
  • Find the probability of observing a sum of 70 or more. To do this, set the lower boundary to 70, and the upper boundary to ∞.

The answer is 3e-05, which means 0.00003 (0.003%).

Extension Question: What is the chance that you observed 40 to 70 heads (inclusive)?

You have one ice cream left at home, and you and your brother desperately want it. Your brother turns to you and says, "Let's play a little game. I want you to roll a 6-sided die 50 times. Every time you roll it, write down the face that the die landed on, and at the end, add up the 50 numbers you recorded rom the 50 rolls. If the sum is greater than 150, then you can have the ice cream, and otherwise, I get to eat it." You agree to his little game.

To do this:

  • In the 'Box Parameters' mini-section, set the tickets to be the numbers 1 through 6, and the number of draws to be 50. Ensure that we are taking the sum for the samples (your brother asked you for the sum after all).
  • Verify that n = 50 is a sufficient number of draws for the central limit theorem to apply for this box model.
  • Find the probability of observing a sum of 150 or more. To do this, set the lower boundary to 150, and the upper boundary to ∞.

We won't give you the EXACT answer this time, but it should be above 90%.

Extension Question: For the game to be fairer, there should be a 50% change that either you or your brother could win. Instead of 150, what should the sum be changed to?

Next week, you and your brother are in the same predicament. One ice cream left - who gets it? Your brother proposes a new game. You roll a 6-sided die 50 times and take the sums, just as you did last time. If the sum is between 145 and 175 (inclusive), you get the ice cream. Otherwise, your brother gets it. What is the probability that you get the ice cream?


Box Model Playground

Box Parameters

First we need to create the box. Follow steps 1, 2, and 3 to configure the box to be whatever you wish to model. The box will output on the right according to the settings that you pick.


First, we need to specify the tickets that we will be adding to the box.

In the text box below, enter the tickets that you wish to place into the box. Only numbers can be added, and each ticket should be seperated by a comma. For example, if you want to have one '1' ticket and one '0' tickets in the box, enter 1,0.

Error: One or move values that you added cannot be interpreted. Please carefully check what you entered. You must enter at least 2 valid tickets. Setting contents of the box to 1,0,0,0.

Second, we need to specify the number of draws (with replacement) that we will be taking from the box:

Error: The value for the 'number of draws' must be an integer greater than or equal to 1. Setting value to 25 until the error is resolved.

Third, we need to specify whether we will be representing the sample using the mean or sum of the draws:




Central Limit Theorem










Modelling Using a Normal Distribution



Finding Probabilities

Lower Boundary

Upper Boundary