
compute a standard deviation, explain how an outlier affects the mean, median and standard deviation, and argue which summary fits a given distribution.
Instruction
Statistics can mislead without a single wrong calculation, if the summary does not fit the data. This lab is about choosing, and defending, the right summary.
Standard deviation. The most widely used measure of spread squares each deviation instead of taking its absolute value. For the data 4, 6, 7, 9, 14:
- The mean is 40 ÷ 5 = 8.
- The deviations are −4, −2, −1, 1, 6.
- The squared deviations are 16, 4, 1, 1, 36, which sum to 58.
- Divide by the number of values: 58 ÷ 5 = 11.6. This is the variance.
- Take the square root: √11.6 ≈ 3.41. This is the standard deviation.
Squaring puts the measure back in the original units only after the square root, and it gives large deviations extra weight: the single value 14 contributes 36 of the 58. Calculators and spreadsheets often report a slightly different version that divides by n − 1 instead of n (here √(58 ÷ 4) ≈ 3.81). The n − 1 version is used when the data are a sample from a larger population. Either is acceptable if you say which one you used.
The outlier problem. Seven apartments in one neighborhood list monthly rents (illustrative): $1,800, $1,950, $2,100, $2,200, $2,300, $2,400 and one luxury unit at $9,500. The mean is about $3,179. The median is $2,200. Without the luxury unit, the mean of the other six is $2,125 and the median is $2,150.
One value moved the mean by more than $1,000 but moved the median by only $50. The standard deviation is even more sensitive: about $2,588 with the outlier and about $204 without it. The mean and standard deviation use every value’s size; the median and IQR use only positions in the ordered list, so they are resistant to extreme values.
The rule of thumb. For roughly symmetric distributions without outliers, report the mean and standard deviation. For skewed distributions or data with outliers, report the median and IQR. This is why news reports on incomes and home prices usually give the median.
But do not delete outliers to make the numbers look nicer. An outlier might be a data-entry mistake, which you should correct, or a real value, which you should report and explain. The luxury unit is real: removing it would hide something true about the neighborhood. The honest move is to report the median and IQR as the summary and mention the outlier in words.
Discourse protocol. Your group receives a claim, such as “The average rent here is $3,179, so most families need $3,000 a month.” One student explains what is true in the claim; another explains what is misleading; a third proposes a better summary. Use sentence frames: “I agree that ___ because ___.” “The claim assumes ___, but the data show ___.” “A better summary would be ___, because the distribution is ___.” Then switch roles for the next claim.
📖 Read for this lesson
The reading opens over the lesson; close it to come back to exactly where you were. The checks below draw on it.
Formative check
Work through these before moving on. They are not graded — they tell you whether the standard below has landed yet.
A $9,500 rent is added to six rents between $1,800 and $2,400. What happens?
Which pair of statistics best summarizes household incomes in a county, where a few households earn far more than the rest?
For 4, 6, 7, 9, 14 the squared deviations sum to , so the standard deviation (dividing by n) is about .
A classmate says: “The outlier makes the data messy, so we should delete it before we compute anything.” Write a response that agrees with one part of this idea and challenges another, using the rent example.
How confident are you that you can decide which measures of center and spread fit a distribution and defend that choice?
Practice
Work these on paper or in your notebook, then open Check your answer. Aim for all of Fluency and Application; try at least one Challenge.
Printable version: this unit’s practice workbook (PDF)
Fluency
Build speed and accuracy with the core skill.
- For the data 4, 6, 7, 13, 17, 29, 37: find the mean, median, range and interquartile range. (Find the quartiles as the medians of the lower and upper halves, leaving out the overall median when the count is odd.)
Check your answer
Answer: mean = 16.14; median = 13.0; range = 33; IQR = 23.0 (Q1 = 6.0, Q3 = 29.0) - For the data 4, 5, 18, 18, 18, 22, 33, 34: find the mean, median, range and interquartile range. (Find the quartiles as the medians of the lower and upper halves, leaving out the overall median when the count is odd.)
Check your answer
Answer: mean = 19.00; median = 18.0; range = 30; IQR = 16.0 (Q1 = 11.5, Q3 = 27.5) - For the data 7, 10, 13, 20, 28, 31, 39, 40, 40: find the mean, median, range and interquartile range. (Find the quartiles as the medians of the lower and upper halves, leaving out the overall median when the count is odd.)
Check your answer
Answer: mean = 25.33; median = 28.0; range = 33; IQR = 28.0 (Q1 = 11.5, Q3 = 39.5) - Find the standard deviation (divide by n) of 2, 4, 4, 4, 5, 5, 7, 9, to two decimal places.
Check your answer
Answer: 2Mean 5; average the squared deviations, then take the square root. - Find the standard deviation (divide by n) of 10, 12, 14, to two decimal places.
Check your answer
Answer: 1.63Mean 12; average the squared deviations, then take the square root. - Find the standard deviation (divide by n) of 1, 3, 5, 7, 9, to two decimal places.
Check your answer
Answer: 2.83Mean 5; average the squared deviations, then take the square root. - Find the standard deviation (divide by n) of 20, 20, 20, 20, to two decimal places.
Check your answer
Answer: 0Mean 20; average the squared deviations, then take the square root.
Application
Use the skill in context. Show your reasoning.
- Annual salaries at a small business, in thousands of dollars (illustrative): 38, 41, 42, 45, 47, 250. Find the mean and median. Which better describes a typical employee, and why?
Check your answer
Answer: Mean 77.17 thousand; median 43.5 thousand. The median, because one salary of 250 thousand skews the data right.An extreme value pulls the mean toward it; the median is resistant. - The data 4, 6, 7, 9, 14 have mean 8 and standard deviation about 3.41. If 5 is added to every value, what are the new mean and standard deviation?
Check your answer
Answer: Mean 13; SD 3.41 (unchanged).Shifting every value moves the center but not the distances between values. - If every value in 4, 6, 7, 9, 14 is doubled, what are the new mean and standard deviation?
Check your answer
Answer: Mean 16; SD 6.81.Doubling doubles every deviation, so both the mean and the SD double.
Challenge
Stretch problems. Expect to think before you write.
- The data are 10, 10, 10, 10 and x. Find every value of x for which the standard deviation (divide by n) is exactly 4.
Check your answer
Answer: x = 20 or x = 0Let d = x − 10. The mean is 10 + d/5; the squared deviations sum to 4d²/25 + 16d²/25 = 4d²/5, so the variance is 4d²/25 and SD = 2|d|/5 = 4, giving |d| = 10.
Review
Keep earlier skills sharp.
- Solve: −7x + 3 = −39
Check your answer
Answer: x = 6Subtract 3 from both sides: −7x = −42. Divide by −7. - Evaluate (−3)² − 4(2).
Check your answer
Answer: 1(−3)² = 9, not −9. - A price rises from $40 to $50. What is the percent increase?
Check your answer
Answer: 25%Divide the change by the original amount.
One Outlier, Two Arguments: The July San Diego Recorded 80.9 Degrees
NOAA National Centers for Environmental Information, Climate at a Glance: City Time Series — July average maximum temperature, Fresno (station USW00093193) and San Diego (station USW00023188), 1995–2025; retrieved September 23, 2026 · Public domain (works of the United States government)
Why this reading. A discourse lab needs a claim worth arguing about and the numbers behind it. Here is one, from the federal climate record: a single July in San Diego that the 1.5 × IQR rule calls an outlier, and two students who want to do different things with it. Every figure you need to judge them is computed below. The argument is not about whether anyone made an arithmetic mistake. Nobody did.
The record
San Diego International Airport, July average maximum temperature, thirty-one Julys from 1995 to 2025. The five-number summary is 70.1, 73.0, 74.1, 75.7, 80.9 °F, so the IQR is 2.7 and the upper fence is 75.7 + 1.5(2.7) = 79.75. One value clears it: 80.9 °F, July 2006, a month in which a severe heat wave crossed the western United States. The reading is real and NOAA has never removed it.
The two positions
Ana: “Eighty-point-nine is flagged as an outlier, so drop it. Then the summary describes a normal San Diego July, which is what the question was about.”
Bo: “Keep it. It happened. Dropping the hottest month from a record of hot months is how you end up telling people July is cooler than it is.”
What each choice actually does to the numbers
| Statistic | All 31 Julys | 30 Julys, 2006 dropped | Change |
|---|---|---|---|
| Mean | 74.28 °F | 74.06 °F | −0.22 |
| Median | 74.10 °F | 73.95 °F | −0.15 |
| Standard deviation (dividing by n) | 2.10 °F | 1.74 °F | −17% |
| IQR | 2.70 °F | 2.30 °F | −0.40 |
| Range | 10.8 °F | 8.5 °F | −2.3 |
Check the mean yourself: the thirty-one values add to 2302.8, and 2302.8 ÷ 31 = 74.284. Take out 80.9 and 2221.9 ÷ 30 = 74.063.
That is a surprise if you came from the rent example in the mini-lecture, where one luxury unit shifted the mean by more than a thousand dollars. Here the hottest July on record moves the mean by about a fifth of a degree. The difference is n. One value out of seven carries a seventh of the weight; one value out of thirty-one carries a thirty-first.
The same value in a small data set
Take only the seven most recent Julys in San Diego: 73.3, 74.4, 76.0, 72.2, 74.1, 76.7, 73.3. Mean 74.29, median 74.1, standard deviation 1.47. Now imagine the 2025 reading of 73.3 had instead been 80.9:
| Statistic | Seven ordinary Julys | Same seven with 80.9 |
|---|---|---|
| Mean | 74.29 °F | 75.37 °F |
| Median | 74.10 °F | 74.40 °F |
| Standard deviation | 1.47 °F | 2.66 °F |
The mean moves five times as far as it did in the thirty-one-value set, and the standard deviation nearly doubles. The median moves 0.3 either way. So “outliers pull the mean” is true, but how hard they pull is a question about sample size, and a claim that ignores n is only half an argument.
What the agency itself does
NOAA does not delete the value. It publishes the monthly figure and, separately, publishes records and rankings so that an unusual month can be identified rather than hidden. That is a third position neither Ana nor Bo took: report the resistant summary and name the outlier in words. “Median 74.1 °F, IQR 2.7 °F, with one July (2006) at 80.9 °F during a regional heat wave” is shorter than this paragraph and loses nothing.
For the protocol
The claim on the table is: “San Diego Julys average about 74 degrees, so the 2006 reading should come out before we summarize.” One of you says what is true in it. One says what is misleading. One proposes a better summary and says which statistic it uses and why. Then argue the second claim: “Removing the outlier barely changed the mean, so it does not matter whether we remove it.” That one is harder, because the first half is a fact from the table above and the second half does not follow from it.
Sources: NOAA NCEI, Climate at a Glance: City Time Series (Fresno) (ncei.noaa.gov); NOAA NCEI, Climate at a Glance: City Time Series (San Diego) (ncei.noaa.gov)