EC 285 - 2026-10-08 - Lecture 09

EC 285 - Lecture 09 - 2026-10-08

Introductory Statistics · Zara Liaqat · SB 106

Week 5 · 16:02-17:19 · 76 min · Full transcript · Flashcards

Overview

This lecture finished Chapter 4 (Discrete Random Variables). It covered the last named discrete distribution, the Poisson, which counts how many times an event happens in a fixed stretch of time. The rest of the class was the portfolio analysis problem promised in earlier lectures. That problem uses expected values, variances and covariance from a probability table to find the average and the risk of a two-stock portfolio. The Chapter 5 slides (continuous random variables) were not reached, and Chapter 5 starts after reading week.

The Poisson distribution

This is the third distribution on the formula sheet, after the Bernoulli and binomial from last lecture.

What the Poisson measures

The Poisson distribution gives the probability of a number of successes (events) happening over a given continuous interval, usually a period of time. Class examples were the number of letters in a mailbox in a week, students emailing the lecturer in the next hour, customers taking cash from an ATM, and customers renting a carpet cleaner.

Why: It lets a business plan for demand. A bank can keep enough cash in the ATM, and a store can stock enough carpet cleaners so customers are not turned away.

Poisson assumptions

Poisson probability formula

The probability of exactly x successes in the interval, given the average rate λ.

Why: Plug in two numbers from the question and you are done. Here λ comes from past data (for example, "on average I get three emails an hour"), and x is the count you are asked about.

P(x)=e−λλxx!

where x is the number of successes, λ is the expected number of successes per unit of time, and e≈2.718 is the base of the natural log.
On the formula sheet for: Midterm 2, final.

Example: Two successes when λ = 0.5

Find P(X=2) when λ=0.5.

  1. P(2)=e−0.5(0.5)22! -- put the numbers into the formula
  2. =0.6065×0.252 -- work out e−0.5, 0.52 and 2!=2
  3. =0.15162=0.0758 -- divide

Answer: P(X=2)=0.0758 -- there is about a 7.6% chance of exactly two events when you expect half an event per period.

Example: Textbook emails, exactly two in a day (λ = 7)

A person receives an average of λ=7 emails per day. Find the probability of exactly 2 emails in a day.

  1. P(2)=e−7722! -- put the numbers into the formula
  2. =0.000912×492 -- work out e−7, 72 and 2!
  3. =0.04472=0.0223 -- multiply, then divide

Answer: P(X=2)≈0.022 -- two emails is very unlikely on a day when seven are expected.

Example: Textbook emails, at most two in a day (λ = 7)

Same setup. "At most two" means 0, 1 or 2 emails, so add the three probabilities, like the addition rule ("this or this or this").

  1. P(0)=e−7700!=0.000912 -- zero emails (70=1 and 0!=1)
  2. P(1)=e−7711!=7×0.000912=0.006383 -- one email
  3. P(2)=0.022341 -- two emails, from the last example
  4. P(X≤2)=0.000912+0.006383+0.022341=0.0296 -- add the three

Answer: P(X≤2)≈0.0296, which the class gave as 0.029 -- about a 3% chance of two or fewer emails in a day.

Exam

"I could ask you a question on this in three different ways": find the expected value, the variance, or the probability (or more than one of these). For the probability, "all you have to do is just substitute everything in the formula... You just need to make sure you're using the calculator correctly, and you don't make a mistake there."

Poisson mean and variance

The mean of a Poisson variable is λ, and its variance is also λ, so its standard deviation is λ.

Why: The mean is λ by definition, because λ is the expected number of successes. Proving that the variance is also λ needs a proof that the lecturer skipped ("you don't need to worry about it"). This is true for the Poisson only.

E[X]=λVar[X]=λσ=λ

where λ is the expected number of successes per unit of time and σ is the standard deviation.
On the formula sheet for: Midterm 2, final (the mean and variance only; take the square root yourself for the standard deviation).

Example: Variance when λ = 0.5

Find the variance and standard deviation of X when λ=0.5.

  1. Var[X]=λ=0.5 -- the variance is just λ
  2. σ=0.5=0.707 -- the standard deviation is its square root

Answer: The variance is 0.5, "exactly the same as λ", with nothing to calculate. The standard deviation is 0.707.

Example: Textbook emails, standard deviation (λ = 7)

Same setup as above, with λ=7 emails per day.

  1. Var[X]=λ=7 -- the variance equals λ
  2. σ=7=2.65 -- take the square root

Answer: σ=2.65 emails -- the daily count typically lands about 2.65 emails from the average of 7.

Exam

For the expected value, "you don't have to do anything. You just have to look at the question and tell me what the lambda is." The lecturer won't always say "this is your λ": "you may have to read the question. It's the expected number of successes."

Time units must match

The λ and the interval in the question must use the same time unit. If λ is per hour but the question asks about the next minute, convert one of them so both are in minutes or both are in hours.

Why: λ is a rate. A rate per hour gives the wrong count for a ten-minute window unless you scale it.

Made-up example: Emails in ten minutes

Emails arrive at λ=6 per hour. Find the probability of no emails in the next 10 minutes.

  1. λ10 min=6×1060=1 -- scale the hourly rate down to 10 minutes
  2. P(0)=e−1100! -- put the numbers into the formula
  3. =0.3679×11=0.3679 -- work out e−1

Answer: P(X=0)=0.368 -- about a 37% chance of a quiet ten minutes.

Exam

"It has to be minutes for both or hours for both." One of the practice problems the lecturer posts has mismatched units, "so you want to make sure they're the same."

Shape of the Poisson

For λ=0.5, the chart of probabilities slopes down: P(X=0)≈0.6, P(X=2)=0.0758, and by X=6 or 7 the probability is almost zero. As λ gets bigger, the distribution starts to look more like a normal, bell-shaped bar chart. It is always centred around λ, the mean.

Why: The lecturer's reason is that, over the same fixed period, many successes are simply less likely than a few. Mathematically, the x! in the denominator grows fast. (Each step multiplies the probability by λ/(x+1), so the chart slopes down from the start only when λ is below 1. With a bigger λ, the probabilities rise up to around λ and then fall.) Everything depends on λ.

Poisson tables

The slides show a Poisson table: the rows are x (the number of successes), the columns are values of λ, and you read the probability where they meet. At x=2 and λ=0.5, it shows 0.0758, the same as the calculator answer.

Exam

"This table will not be provided... don't rely on the table. For this chapter, for this distribution, you need to be able to do it on the calculator." The table only covers a few values of λ. Binomial tables are not given either (from last lecture). The tables you will get are the normal distribution table and the Student's t-distribution table.

Lambda and mu in the textbook

The textbook's Poisson section writes the average rate as μ instead of λ. The lecturer, the slides and the formula sheet all use λ. Both mean the same thing, the expected number of successes. When you read the book or its solutions, read μ as λ. The lecturer called this confusing, because μ is also used for the mean elsewhere.

Spotting a Poisson question

A Poisson question has an element of time: a number of events "in a day", "in the next hour" or "per week". The lecturer will go over how to tell all the distributions apart once Chapter 5 is done ("remind me").

Exam

"The question will not always tell you which formula to use, so you should know which one." For the Poisson, the time dimension is the clue.

Finishing Chapter 4

Joint distributions review

The last Chapter 4 slides (joint and marginal distributions, conditional probability P(y∣x)=P(x,y)/P(x), and independence P(x,y)=P(x)P(y)) were skipped as "nothing new". They are the probability-table ideas from Lecture 05, applied to random variables. The tutorial after reading week will cover a way to test with these.

Portfolio analysis

This applies the mean and variance of a linear function of random variables, from Lecture 06 and the portfolio preview, with real numbers.

The portfolio setup

The price of stock A is the random variable X, and the price of stock B is the random variable Y. A portfolio holds a of stock A and b of stock B (a number of shares, or percentages like 40/60). Its market value W is a linear function of X and Y, so W is random too.

Why: Prices move with the economy, so you can't know what the portfolio will be worth. You can still work out its expected value and how much it varies, which is what ETFs and other stock baskets are built on.

W=aX+bY

where X and Y are the prices of stocks A and B, and a and b are the amounts held of each.

The class example used these three outcomes:

Economy Probability X (stock A) Y (stock B)
Recession 0.2 −$25 −$200
Stable 0.5 +$50 +$60
Expansion 0.3 +$100 +$350

The weights were a=0.4 (40% in stock A) and b=0.6 (60% in stock B). The question was to find the average return on W and how much it varies. You do it in five pieces: E[X], E[Y], Var(X), Var(Y) and Cov(X,Y), then put them into the portfolio formulas.

Exam

The lecturer is posting the notes for this example "because so many students are not here today and then they'll get a really bad score at the midterm when I ask a similar question." With practice, the whole problem "should take roughly 15–20 minutes."

Expected value from a table

The expected value of a discrete random variable is each value times its probability, all added up. This is the same method as the two-coin example in Lecture 06.

Why: It is a weighted average. Likelier outcomes count for more. It is "expected" because we don't know how the economy will actually turn out.

E[X]=μX=∑xP(x)

where the sum runs over every possible value x, and P(x) is its probability.
Not on any formula sheet, so memorise it.

Example: Expected price of stock A

Use the table: X is −25, 50 or 100 with probabilities 0.2, 0.5 and 0.3.

  1. E[X]=0.2(−25)+0.5(50)+0.3(100) -- multiply each value by its probability
  2. =−5+25+30 -- work out each term
  3. =50 -- add them

Answer: E[X]=$50 -- the average return on stock A.

Example: Expected price of stock B

Y is −200, 60 or 350 with probabilities 0.2, 0.5 and 0.3.

  1. E[Y]=0.2(−200)+0.5(60)+0.3(350) -- multiply each value by its probability
  2. =−40+30+105 -- work out each term
  3. =95 -- add them

Answer: E[Y]=$95 -- stock B earns more on average than stock A.

Mean of a portfolio

The expected value of W=aX+bY is the same weights applied to the two expected values.

Why: Expected values pass straight through a linear function. Each stock contributes its share of its own average.

E[W]=aE[X]+bE[Y]

where a and b are the portfolio weights.
Not on any formula sheet, so memorise it.

Example: Average return on the portfolio

With E[X]=50, E[Y]=95, a=0.4 and b=0.6:

  1. E[W]=0.4(50)+0.6(95) -- put the numbers into the formula
  2. =20+57 -- multiply
  3. =77 -- add

Answer: E[W]=$77 -- the average return on the 40/60 portfolio.

Variance from a table

The variance of a discrete random variable is the squared distance of each value from the mean, times its probability, all added up. It is the usual "squared deviations from the average", weighted by probability.

Why: Squaring stops positive and negative gaps from cancelling out, and weighting by probability makes likelier outcomes count for more.

σX2=E[(X−μX)2]=∑(x−μX)2P(x)

where μX is the mean of X, found first, and P(x) is the probability of each value.
Not on any formula sheet, so memorise it.

Example: Variance of stock A

Use μX=50 in every term. There are three outcomes, so there are three terms.

  1. σX2=(−25−50)2(0.2)+(50−50)2(0.5)+(100−50)2(0.3) -- subtract the mean from each value
  2. =(5625)(0.2)+(0)(0.5)+(2500)(0.3) -- square each gap
  3. =1125+0+750=1875 -- multiply by the probabilities and add
  4. σX=1875=43.30 -- take the square root for the standard deviation

Answer: Var(X)=1875 and σX=$43.30 -- how much stock A's return swings around its $50 average.

Example: Variance of stock B

Use μY=95 in every term.

  1. σY2=(−200−95)2(0.2)+(60−95)2(0.5)+(350−95)2(0.3) -- subtract the mean from each value
  2. =(87025)(0.2)+(1225)(0.5)+(65025)(0.3) -- square each gap
  3. =17405+612.5+19507.5=37525 -- multiply by the probabilities and add
  4. σY=37525=193.71 -- take the square root

Answer: Var(Y)=37525 and σY=$193.71 -- stock B is far more volatile than stock A. The lecturer's notes first said 193.21, and a student corrected it to 193.71.

Exam

"Better do this now than do it the day before the exam."

Covariance of two random variables

Covariance measures whether two random variables move together. For a probability table, multiply each outcome's two deviations from their means by that outcome's probability, then add them up.

Why: A positive covariance means the two stocks tend to rise and fall together. Its size depends on the units (here dollars), so a big number is hard to judge on its own. That is why correlation, which is always between −1 and 1, is the easier next step. The lecturer didn't ask for correlation in this problem.

Cov(X,Y)=E[(X−μX)(Y−μY)]=∑(x−μX)(y−μY)P(x,y)

where μX and μY are the means, and P(x,y) is the probability of each outcome (row of the table).
The lecturer said this "is given in the formula sheet". The Midterm 1 and Midterm 2 sheets only print the population covariance for a data set (dividing by N), not this probability-weighted version, and the final sheet has no covariance. So memorise this form.

Example: Covariance of the two stocks

Use μX=50 and μY=95. Each row of the table gives one term.

  1. Cov(X,Y)=(−25−50)(−200−95)(0.2)+(50−50)(60−95)(0.5)+(100−50)(350−95)(0.3) -- take each row's two deviations
  2. =(−75)(−295)(0.2)+(0)(−35)(0.5)+(50)(255)(0.3) -- work out the deviations
  3. =(22125)(0.2)+0+(12750)(0.3) -- multiply the pairs
  4. =4425+0+3825=8250 -- multiply by the probabilities and add

Answer: Cov(X,Y)=8250 -- a strong positive co-movement, so the two stocks rise and fall together.

Exam

"Everybody should work on this, even if you're not in the mood to do that right now... definitely solve it for the exam."

Variance of a portfolio

The variance of W=aX+bY adds the two weighted variances plus a covariance term.

Why: The lecturer compared it to expanding (a+b)2: two squared terms plus 2ab times the cross term. The covariance term captures whether the two stocks' swings add up or cancel out.

σW2=a2σX2+b2σY2+2abCov(X,Y)

where a and b are the weights, σX2 and σY2 are the variances of X and Y, and Cov(X,Y) is their covariance.
Not on any formula sheet, so memorise it.

Example: Variance of the portfolio return

Use a=0.4, b=0.6, σX2=1875, σY2=37525 and Cov(X,Y)=8250.

  1. σW2=(0.4)2(1875)+(0.6)2(37525)+2(0.4)(0.6)(8250) -- put the numbers into the formula
  2. =0.16(1875)+0.36(37525)+0.48(8250) -- square the weights and work out 2ab
  3. =300+13509+3960=17769 -- multiply and add
  4. σW=17769=133.30 -- take the square root

Answer: Var(W)=17769 and σW=$133.30 -- the portfolio's typical swing around its $77 average. The lecturer's notes said 133.04, but a student's 133.30 was accepted, and 133.30 is correct.

Risk versus return

Stock B has the higher expected return ($95 against $50), but also a much bigger spread (standard deviation $193.71 against $43.30). Which one to favour "depends on your personality, how risk averse, risk lover you are." Someone who wants stability puts more in stock A (a bigger than b). Someone who likes risk puts more in stock B for the big reward if the economy does well. A student pointed out that the probabilities come from past data and could be off. The lecturer agreed: that's why it's "expected", not actual.

Exam checklist

Admin & deadlines


Transcribed automatically from the lecture recording. Slides used: EC 285 - Ch. 05 - Continuous Random Variables (slides).pdf. Course info used: EC 285 - Formula Sheet - Final.pdf, EC 285 - Formula Sheet - Midterm 1.pdf, EC 285 - Formula Sheet - Midterm 2.pdf, EC 285 - Syllabus - Fall 2026.pdf. Audio archived at /mnt/porsche/configs/lectures/archive/2026-10/EC 285 - 2026-10-08 - Lecture 09.opus.