EC 285 - 2026-09-15 - Lecture 02
EC 285 - Lecture 02 - 2026-09-15
Week 2 · 15:59-17:17 · 77 min · Full transcript · Flashcards
Overview
This lecture finished Chapter 1 and started Chapter 2. It covered why we sample, the two sampling methods you are responsible for, descriptive versus inferential statistics, the types of variables, and how a histogram differs from a bar chart. Chapter 2 then began with the two big families of summary numbers, centre and spread. The lecture worked through the mean, median and mode with a practice problem, and ended on percentiles and quartiles. Variance, standard deviation, covariance and correlation are on the Chapter 2 slides but were not reached yet.
Why we sample
This finishes the Chapter 1 slides. Part of it reviews Lecture 1, and it uses the same terms (population, sample, representative).
Why we need a sample
Statistics gives us tools to process (collect), summarize, analyze and interpret data. We collect data from a sample and use it to learn about the larger population. We can't just survey the whole population because it takes too much time and money. Often it isn't even possible, because of non-response: many people ignore survey requests. The lecturer said that when she is busy she deletes survey emails without opening them. A small chance to win a $25 gift card doesn't change that for her, but a guaranteed one does.
Why: Since we only ever see part of the population, any claim we make about the population stays uncertain, even with a good sample.
Representative samples
A representative sample looks like the larger population. For example, it should not be all very old people, all middle-aged people or all women. A bigger sample helps, but being representative matters more. Statistics Canada makes its samples representative by stratifying. It sets shares in advance (for example 20% from smaller provinces or rural areas) and then picks at random within each group.
Why: If the sample is lopsided, your claims about the population are lopsided too, however big the sample is.
Stratified sampling was only described as background. The lecturer said "just focus on these two" (simple random and systematic) and "ignore the other method that is in the book." "You're responsible for only these two types of sampling methods."
Simple random sampling
Individuals or items are picked from the population strictly by chance.
- Any group of
individuals is equally likely to be chosen as any other group of individuals.
Why: Pure chance means no group gets favoured on purpose, so the sample tends to be representative.
Systematic sampling
First, arrange the population in an order that has nothing to do with what you are studying. (For example, if you're trying to study about how age affects something, don't arrange the population by age). Then, pick a random starting point among the first J items, and take every Jth item after that. It is more organized than simple random sampling, but the random start keeps some chance in it.
Why: The random start keeps the selection chance-based. The fixed step spreads the sample evenly across the whole list. The order must not be related to your question, so if you are studying income, don't sort by income.
where N (capital) is the population size and n (lowercase) is the sample size.
Not on any formula sheet, so memorise it.
You want a sample of
-- put in the sizes -- so take every 8th item - Pick a random number from 1 to 8: say
-- this is the only random step -- keep adding 8
Answer: The sample is items 3, 11, 19, 27, 35, 43, 51, 59 and 67. That makes nine items, and the last one is 67.
Simple random sampling and systematic sampling are the only two sampling methods you are responsible for. The lecturer called this 72-and-9 example "the best one to understand how it works."
Descriptive and inferential statistics
Two branches of statistics
- Descriptive statistics summarizes the data you have. That means all the charts (bar charts, pie charts and so on) plus summary numbers like the mean, median and mode. It is Chapters 1 and 2, and it is the first thing you do with new data.
- Inferential statistics uses the sample to test hypotheses about the population. One example is running a regression to test whether Canadians spend less on entertainment when gas prices rise. The lecturer placed this in Chapters 9 and 13, the last part of the course. Inference continues in follow-up courses such as EC 295 and EC 395.
- In between is probability theory: theorems and probability distributions that prepare you for inference.
Why: Description tells you what your sample looks like. Inference is how you get from the sample to a claim about the whole population, which is the goal of the course.
"It's a math course, not an economics course." Exams will have "numerical questions where you have to find the probability and the area under the graph." She also said: "If you want to do well on exams, you should know every single thing that's mentioned on the slides." But "I'm not going to test you on something that wasn't part of the lectures." She also said it is important to keep comprehensive notes to go with the slides.
Types of variables
Discrete and continuous variables
- Discrete numerical variable: takes whole numbers only, never decimals. Examples are the number of classes you take (5, not 5.5) and the number of children. (This is the lecturer's rule of thumb. More precisely, a discrete variable takes separate, countable values, and these are usually counts. Her whole-number rule works for all of her examples.)
- Continuous numerical variable: can be any value on the number line, including decimals. It is usually measured over a range. Every numerical variable that isn't discrete is continuous.
Why: The type of variable decides which charts and summaries make sense.
"Lots of new terms in this chapter, and you should know all of them." "Anything that's in bold is especially important... I might ask multiple-choice questions on these terms." You don't have to recite definitions word for word, but you need to be able to pick the right option.
Categorical variables
A categorical variable sorts things into groups instead of measuring a number. Examples are hospital units, gender, relationship status, and hair or eye colour.
Why: You can't average categories, so they get their own charts and summaries.
Charts for each type
Bar charts and pie charts are the most common charts for categorical variables. Numerical variables have more options, including the histogram. Bar charts can show numerical variables too.
Only study the chart types listed on the slides. Skip the box-and-whisker plot and the other extra charts in the book: "I won't be testing you on those."
Frequency distributions and histograms
Frequency distribution table
This is a two-column table. The first column lists the values (for example 0, 1, 2, 3... children). The second column shows the frequency, meaning how many observations have each value (for example how many families have no children, one child, and so on). In this example both columns are discrete, because you can't have 5.5 instances. When you group values into ranges like 10 to 20 and 20 to 30, those ranges are the class intervals, and their edges are the class boundaries.
Why: For long data sets, a table or histogram is much easier to read than a raw list. The lecturer said it is also important when calculating averages.
Histogram versus bar chart
In a histogram, the area of each rectangle is proportional to the frequency. Interval widths don't have to be equal. You can merge empty or sparse intervals into one wider bar, as long as you lower its height so the area still matches the frequency. In a bar chart every bar has the same width, and only the height changes. Histograms also need numerical data, and the x-axis has to be in order. You can't draw a histogram for a categorical variable. Some answers from class did not count as the difference, because a bar chart can have them too: no gaps between the bars, multiple bars, and a variable on each axis.
Why: Because area carries the frequency, the researcher is free to choose the intervals without distorting the picture.
The intervals are 10 wide. The lecturer merges 20 to 30 and 30 to 40 into a single bar from 20 to 40 that holds a frequency of 11.
-- two intervals merged into one -- divide by 2 so the area stays at 11 -- area (in old-width units) matches the frequency
Answer: The bar's height is 5.5, not 11. The bar is twice as wide, so its area still equals the frequency of 11.
This got a long discussion in class, and the lecturer pointed out that "this is not explained in the book." The phrase she was looking for is "the area of the rectangle that you end up with should be proportional to the frequency."
Describing data numerically
This starts Chapter 2. Everything here is descriptive statistics.
Centre and spread
The first two things to check in any data set are where the centre is (measures of central tendency) and how spread out the data is around it (measures of variation). The lecturer showed a bell-shaped frequency curve of midterm grades: grade on the x-axis, how often each grade occurs on the y-axis. It is smooth because grades there are treated as a continuous variable. Whether a small spread is "good" depends on the centre. A small spread around a 70 average is ideal. Around a 40 average it would mean most people failed. In that case she would rather see a bigger spread around the low average. Around a 90 it might mean the exam was too easy. The lecturer said one line in Stata (summarize grade, or sum grade) gives the mean, median, minimum, maximum and other summaries in an instant. (In Stata itself, plain summarize shows the count, mean, standard deviation, minimum and maximum. You need summarize grade, detail to get the median and percentiles, and neither version shows the mode. This only matters for the assignments, since there are no Stata questions on exams.)
Why: The average alone hides how typical it is. Two data sets can share a centre and have very different spreads.
"There will be questions on the first exam where I ask you to find these numbers or define them." Midterm 1 is October 4, 12:00 to 1:00 PM (from the syllabus).
Measures of variation
There are five: range, interquartile range, variance, standard deviation and coefficient of variation. They describe how spread out the data values are. Their formulas are on the slides but were not covered yet.
Why: They tell you whether a given value is close to or far from the centre.
"You should be familiar with all five of them." She called variance and standard deviation "important terms," and said "we will talk about these throughout the course. For every probability distribution theorem, there is going to be a formula for variance and standard deviation."
Measures of central tendency
Mean
The arithmetic mean is the average: the sum of all observations divided by the number of observations.
Why: It uses every value, so it is the natural "typical value." That same fact is why extreme values pull it around.
where
Not on any formula sheet, so memorise it.
A population of seven companies has 1, 2, 2, 3, 4, 4, 5 employees.
-- add all values, divide by the count -- the sum is 21 -- divide
Answer: The mean is 3 employees.
Median
The median is the middle value once the data is sorted in ascending order. It is also called the 50th percentile, because half the observations are below it and half are above it. With an even number of observations, take the average of the two middle values.
Why: Outliers and extreme values move the mean but not the median. That makes the median useful when data is skewed or has outliers.
Same seven companies: 1, 2, 2, 3, 4, 4, 5. The data is already sorted.
is odd, so there is one middle value -- the 4th - The 4th value is
-- count 1, 2, 2, 3
Answer: The median is 3, the same as the mean here.
Add an eighth company with 7 employees: 1, 2, 2, 3, 4, 4, 5, 7.
is even, so there are two middle values -- the 4th and 5th - The middle values are
and -- average them
Answer: The median is 3.5.
For an even count, the lecturer will accept 3.5, and "if you say it's 3 and 4, I would consider that correct as well." (3.5 is the standard answer, so give that one on a test.)
Outliers and shifts
Changing an extreme value moves the mean but not the median. Adding the same amount to every value moves both the mean and the median up by that amount.
Why: The median only depends on which value sits in the middle position. The mean depends on every value.
Change the last company's 5 employees to 10: 1, 2, 2, 3, 4, 4, 10.
- The median is still the 4th value:
-- the middle position hasn't changed -- the sum rises from 21 to 26 -- divide
Answer: The median stays at 3, and the mean goes up from 3 to about 3.71.
A recording error means every company actually has 3 more employees: 4, 5, 5, 6, 7, 7, 8.
-- the mean was 3 - The median is the 4th value:
-- it was 3
Answer: Both go up by exactly 3. In the same way, giving everyone 2 bonus marks would raise the class average by 2 ("don't count on that").
The lecturer said this practice problem shows "how mean, median would be impacted if there is outliers, if there's a mistake in value itself." The outlier question has two answers: the median does not change, and the mean goes up.
Mode
The mode is the most frequently observed value. Data may have no mode if every value appears only once.
Why: It shows the most common value, and it works even for categories.
Shape and skewness
Real data is rarely perfectly symmetric. It is often skewed, meaning it has a long tail on one side.
- Left-skewed: mean < median (the tail pulls the mean down)
- Symmetric: mean = median
- Right-skewed: median < mean
Why: Comparing the mean and median tells you which way the data leans. That is why you report both.
Geometric mean
The geometric mean is another kind of average: the n-th root of the product of the values. It is used for average rates of return and growth (markets, investment, population).
Why: Growth compounds, so rates multiply instead of adding. A plain average of growth rates overstates the true average whenever the rates differ.
where
Not on any formula sheet, but you don't need to memorise it, because it is not tested.
An investment falls 10% in year 1 and rises 60% in year 2, so the growth factors are 0.9 and 1.6.
-- multiply the growth factors -- take the square root, since -- subtract 1 to get a rate
Answer: The average is 20% per year. Check:
This one is not tested: "You don't need to know this formula. I'm not going to test you on geometric means." She suggested putting a cross on that slide.
Percentiles and quartiles
This uses the same seven-company data. The median from the previous section turns out to be a special percentile.
Percentiles
The P-th percentile is the value that P% of the observations fall below, with the rest above. For example, an IQ at the 90th percentile means 90% of people score lower and 10% score higher. Percentiles are generally used for large data sets.
Why: A percentile shows where a value stands relative to the whole data set, not just the average.
where P is the percentile you want and n is the number of observations. The answer is the value sitting in that position once the data is sorted in ascending order.
Not on any formula sheet, so memorise it.
Data: 1, 2, 2, 3, 4, 4, 5, with
-- put in P and n -- we want the 6th position - Count 1, 2, 2, 3, 4, 4: the 6th value is
-- read off the sorted data
Answer: The 75th percentile is 4 employees, not 6. Six is only the position.
"This is not the answer, and this is the mistake a lot of students make. The 75th percentile is not 6." "This formula here tells you what position you're looking at, not the final answer."
Quartiles
Quartiles are the special percentiles that split sorted data into four parts, each with an equal number of values. The parts can have different widths.
Why: They are the most widely used percentiles, and they also give you the interquartile range later on.
Same data: 1, 2, 2, 3, 4, 4, 5, with
-- we want the 2nd position - The 2nd value is
-- count 1, 2 - Check
: , and the 4th value is 3 -- this matches the median
Answer:
Exam checklist
- #Representative samples: stratified sampling was only background; ignore the book's other methods.
- #Systematic sampling: J = N/n, random start, then add J each time.
- #Two branches of statistics: a math course; everything on the slides is fair game, nothing outside lectures.
- #Discrete and continuous variables: bold Chapter 1 terms may come up as multiple-choice questions.
- #Charts for each type: only the chart types on the slides; skip box-and-whisker.
- #Histogram versus bar chart: area is proportional to frequency.
- #Centre and spread: Midterm 1 asks you to find and define these measures.
- #Measures of variation: know all five; variance and standard deviation come back all term.
- #Median: for an even count, 3.5 or "3 and 4" are both accepted.
- #Outliers and shifts: an outlier moves the mean, not the median; a shift moves both.
- #Geometric mean: not tested.
- #Percentiles: the formula gives the position, not the answer.
Admin & deadlines
- Tutorials start this week. Tutorial 3 is Thursday 7:00 to 8:20 PM, and Tutorials 1 and 2 are Friday at 8:30 and 10:00 AM, all in LH1011. Only go to the tutorial and lecture section you are enrolled in. This is EC 285 Section B, Tuesday and Thursday 4:00 to 5:20 PM, SB106. She said they will be stricter from now on about students attending only their own section. In her busiest section, some students couldn't find a seat because students from other sections had taken spots.
- The first three tutorials are Stata labs (this week and the next two). Bring a laptop with Stata working if you can. The IAs will project their screens either way. The Lab 1 notes are already posted under Stata labs on MyLS. The middle of the term has regular problem-solving tutorials, and there are three or four more Stata labs before Assignment 2.
- Tutorial mini-quiz: it is not about Stata. It is an easy-to-medium Chapter 2 question in two parts with calculations, based on today's or Thursday's lecture. Hand in a sheet with your name and ID even if it's blank, because it doubles as attendance. A blank sheet gets half the mark. Together, submission and correctness are worth one point per tutorial. The syllabus words it as "0.5 mark for submission; 1 mark for correctness," and your two lowest tutorial marks are dropped. (Read literally, that syllabus line could mean 1.5 marks in total. The lecturer said one point in class, so ask her if the difference matters to you.)
- Readings: Chapters 1 and 2 by the end of this week. Only read the sections covered in class or listed on each chapter's cover slide. She suggested saving a PDF of the textbook so you have it offline.
- Assignment 1 covers descriptive statistics in Stata. You will likely choose one categorical and one numerical variable and draw charts for them. Assignment 2 covers inference and is due December 4. Stata is only for the assignments (15%). There are no Stata questions on exams. Don't buy Stata: you have free access, including in the campus labs.
- Midterm 1: October 4, 12:00 to 1:00 PM, worth 20% (from the syllabus). October 4, 2026 is a Sunday, so it's worth confirming the date on MyLS.
- Next class (Thursday, September 17): another practice question at the start, and a worked frequency-distribution example with class intervals. Review your notes before class.
- Room: the lecturer asked for a room where everyone can see both the slides and the board, but no change has happened yet. Sit toward the centre so you can see the board.
- Emails: You should have two emails from the lecturer by now. One is the welcome notes (also posted on MyLearningSpace). The other is about Stata setup and tutorials.
Transcribed automatically from the lecture recording. Slides used: EC 285 - Ch. 02 - Descriptive Statistics (slides).pdf. Course info used: EC 285 - Syllabus - Fall 2026.pdf. Audio archived at /mnt/porsche/configs/lectures/archive/2026-09/EC 285 - 2026-09-15 - Lecture 02.opus.