ETC1000 Cheatsheet · Business and Economic Statistics

Monash University · Cheatsheet for the final exam. Monash Semester 2 final assessment period, 2–18 November 2026 (check your personal exam timetable). Full ETC1000 study guide →

A4 landscape · 7pt · 6 columns · ink + yellow · revision sheet2 sides · ETC1000 · Monash
ETC1000
Business and Economic Statistics
Monash University · Undergraduate Level 1 · Semester 2, 2026
ETC1000 cheatsheet QR code – open the full ETC1000 study guide
Full guide + ask Sia
about ETC1000
SIDE 1/2   DEFINITIONS → FORMULAS → DECISIONS → DEFENSIBLE SENTENCESEXAM REVISION SHEETExamination 40% · end-of-semester period

0 · Exam Mapcourse.json · brief

Exercises 20Tests 20Project 20Examination 40
Exercises, completed during your allocated tutorial20%
In-Class Tests, two tests worth 10 per cent each20%
Group Project20%
Examination40%

Permitted materials: not confirmed for this offering in the course materials we hold — check the exam instructions on the LMS. BRIEF.md

The unit’s repeated scoring move is quantity + sentence: compute the right object, name its denominator or scale, then state what it establishes and what it does not.

1 · Variable → Display → SummaryCh 1–2

WHAT DOES THE VARIABLE RECORD?GROUP / KINDcategoricalHOW MUCH / MANYnumericalnominal / ordinaldiscrete / cont.counts · mode · barcentre · spreadhistogram · boxplot
F1.1 redraw: classify first; every valid display and summary follows.
TypeTest / exampleUse
Nominalno order; paymentcounts, mode, bar
Ordinalordered bandsordered bar, median band
Discretewhole countsfrequency, histogram
Continuousmeasured refinementhistogram, boxplot, mean

Digits do not decide type: postcode, customer ID and store number are nominal because an average is meaningless. A sorted bar chart compares lengths and usually beats a pie chart beyond about four categories.

2 · Contingency TablesCh 1

same cell, changing denominatormarginal = row/column total ÷ grand total
joint = interior cell ÷ grand total
conditional = cell ÷ its row/column total

Independence: compute the same conditional probability inside each row. Equal conditionals → independent; unequal → relationship. Report the percentage-point gap. Signal words given, among, of those who demand a row or column denominator.

3 · Centre, Spread, ShapeCh 2

MeasurePromiseTrap
Meanbalance point; uses all valuespulled by extremes
Medianmiddle rankignores distance
Modemost commonmay tie / be absent
Rangemax − mintwo least typical cases
IQRQ3 − Q1; middle halftails omitted
SDtypical spread about meanquote in original units

sample summariesx̄ = Σxᵢ/n
s² = Σ(xᵢ−x̄)²/(n−1) · s = √s²
IQR = Q3−Q1 · fences = Q1−1.5IQR, Q3+1.5IQR

IQR = 30.25 − 22.5 = 7.751822.52630.2534flagflag ≠ error; investigate
F2.1 redraw: box, whiskers and the 1.5-IQR investigation flag.

Skew: mean > median → long upper/right tail; mean < median → long lower/left tail. Prices, incomes and waiting times often need median + IQR. The mean remains the right centre when the question is about a total.

Why n−1? Deviations use the sample mean, which makes them unusually small; dividing sample squared deviations by n would systematically understate population variance. Use n−1 for a sample estimate.

Histogram rule: try two or three interval widths and retain one where the shape is stable. Too wide hides skew; too narrow reports sampling noise as structure. State the width in the caption.

Outlier workflow: flag → check recording and case context → report sensitivity with and without if influential. A flagged point may be the most informative case and is never deleted merely for being unusual.

4 · Standardise Before ComparingCh 2

mean 60value 68SD 4 → z=+2.00SD 10 → z=+0.80
F2.2 redraw: the same raw value has different standing in each group.

5 · Population → IntervalCh 3

Population languageSample language
population: every case the claim coverssample: observed subset
parameter: fixed, usually unknownstatistic: known, varies by sample
bias: frame/process errorsampling variability: random noise

A larger sample shrinks noise, not bias. Selection, survivor and non-response bias remain. Ask how cases entered before asking how many there are. The law of large numbers stabilises sample summaries around population values; it cannot repair the wrong frame.

PROBLEMquestionSAMPLEcollectMODELfitINFERclaim?Is the last arrow earned?
F5.1 redraw: inference is the last, conditional step.

sampling distribution of the meancentre = μ · SE(x̄) = σ/√n ≈ s/√n
quadruple n → halve SE

interval for a population meanx̄ ± t* × s/√n · df = n−1

UseReason
Normal critical valuepopulation σ known
t critical valueσ unknown; estimate with s; heavier tails

95% confidence: the procedure captures the parameter in about 95 of 100 repetitions. The realised interval either covers it or does not. An interval for a mean says nothing about where most individual observations lie.

FailureWhat went wrongDoes larger n fix it?
Selection biasentry relates to measured quantityNo
Survivor biasfailed cases absent from frameNo
Non-responserespondents differ from non-respondentsNo
Sampling noisechance sample-to-sample movementYes, at √n rate

6 · Precision LeversCh 3

larger swider interval
larger nnarrower at √n rate
higher confidencelarger critical value → wider

Check Write the SE first. If it did not shrink when n grew, you divided by n instead of √n. Then name the population supported by the sampling frame.

7 · Association → RegressionCh 4

Scatter first: direction, form, strength and exceptions decide whether a straight-line summary is licensed. Put the explanatory variable on x and outcome on y. Swapping axes changes the fitted line because least squares measures vertical error; it does not change correlation.

joint movementsxy = Σ(xᵢ−x̄)(yᵢ−ȳ)/(n−1)
r = sxy/(sx sy), −1 ≤ r ≤ 1

r=+1.00r=+0.26r=−1.00
F4.2 redraw: r measures linear pattern, not any pattern.

least squaresŷ = b₀+b₁x · b₁ = sxy/sx² · b₀ = ȳ−b₁x̄
eᵢ = yᵢ−ŷᵢ · choose line minimising Σeᵢ²
R² = r² in simple regression

residual −0.4ŷ = 2.26 + 0.86xvertical miss = actual − predicted
F4.1 redraw: seven stores, fitted line and one residual.

Slope sentence: “A one-unit increase in x is associated with b₁ units of y.” Name both units. Intercept only if x=0 is observed and meaningful. Residual curvature → wrong form; fan → changing spread; isolated large residual → influential exception.

R² is not a licence: it gives the share of sample outcome variation accounted for by the line; it proves neither linearity, causation nor validity outside the observed x-range.

12 · R + Wrangling RulesCh 7–8

Verb / objectChangeCheck
filterkeeps rowscondition first
selectkeeps columnsvariables retained
mutateadds columnunits / type
arrangereordersno data change
groupinvisible labelexpected groups
summarisecollapses rows; changes graincarry count

8 · Hypothesis-Test PipelineCh 5

  1. Name the population parameter.
  2. Write H₀ as equality/no effect; H₁ from the question.
  3. Fix α before seeing data; choose one- or two-sided from wording.
  4. Compute test statistic = (estimate − null)/SE.
  5. Compare p with α; decide; report effect, units and limit.

common formsH₀: μ=μ₀; H₁: μ≠μ₀ (differs/changed)
H₀: β₁=0; H₁: β₁≠0
“exceeds / improved / faster” → one-sided

COMPARE p WITH αα fixed in advancep < αreject H₀p ≥ αdo not rejectnever “prove H₀”
F5.2 redraw: the decision is mechanical once p exists.

Correct p sentence: if H₀ were true, a result at least this extreme would occur with probability p. A large p means insufficient evidence, not that H₀ is true. Statistical significance does not certify commercial importance; read effect size and interval.

ErrorMeaningLever
Type Ireject true H₀; false alarmrate = α
Type IImiss a real effectpower = 1−β
Tradelower α usually raises βlarger n improves both

9 · Multiple RegressionCh 6

coefficient sentencechange in y for +1 unit of xⱼ, holding every other included variable fixed

Adding a variable changes the question and may shrink or reverse a coefficient because shared variation is reassigned. It does not make either estimate arithmetically wrong. Avoid causation from observational data and remember that omitted variables can still contaminate the model.

DeviceRuleRead
k-category factork−1 dummieseach vs omitted baseline
Interactionx×ddifference between slopes
Ordinary R²never falls when x addednot model chooser
Adjusted R²penalty per variablecompare specifications

10 · ClassificationCh 9

A model gives an estimated positive probability; a threshold turns it into a decision. Lower threshold → more positives, higher sensitivity and more false positives. Raise it when false action is costly; lower it when a missed positive is costlier. Moving the threshold trades errors; it does not improve the fitted model.

PREDICTEDpositivenegativeACTUALpositivenegative48 TPcaught20 FNmissed12 FPfalse flag120 TNclearedacc .84 · sens .71 · spec .91n=200 · rates use different denominators
F9.1 redraw: four counts before four rates.

ratesaccuracy=(TP+TN)/all
sensitivity=TP/(TP+FN) · specificity=TN/(TN+FP)
precision=TP/(TP+FP) · base rate=(TP+FN)/all

RateDenominator asksLead when…
Accuracyall casesclasses not badly imbalanced
Sensitivityactual positivesmisses are costly
Specificityactual negativesfalse alarms are costly
Precisionpredicted positivesa positive flag must be trusted

Accuracy must beat the base-rate benchmark. At 3% defaults, predicting none scores 0.97 and catches nothing. Fit on training data; judge on held-back data.

11 · Logistic RegressionCh 10

1.5p=.5 tipping pointbounded curve · unbounded line
F10.1 redraw: binary outcomes require a bounded probability curve.

scalesodds=p/(1−p) · log odds=ln[p/(1−p)]
log odds = b₀+b₁x+… · odds multiplier for +1 x = eᵇ¹
p = 1/(1+e⁻η), η=b₀+b₁x+…

https://www.asksia.ai/cheatsheet/monash-etc1000 · Full ETC1000 study guide →
v2 · 2026-10-25
Forward this to your tutorial groupShared by ________
Check your exam instructions on the LMS for what you may bring; use this sheet to revise and self-test.
Monash Semester 2 final assessment period, 2–18 November 2026 (check your personal exam timetable).
ETC1000
Business and Economic Statistics · Self-Test Drills
Worked paths · judgement cards · classic traps · exam-morning recall
ETC1000 cheatsheet QR code – open the full ETC1000 study guide
Full guide + ask Sia
about ETC1000
SIDE 2/2   COVER THE ANSWER → WORK THE PATH → CHECK THE SENTENCESELF-TEST DRILLSFive worked methods · judgement cards · morning plan

13 · Worked 1 — ProbabilitiesCh 1 worked

Data: 1,200 returns. Online 750, in store 450. Refunded: 300 online, 90 in store.

OUTCOMErefundedothertotalonlinein storetotal300450750903604503908101200conditionals: .40 online vs .20 in store
Source-data redraw: choose the denominator from the wording.
  1. Marginal refund: (300+90)/1200 = 0.325.
  2. Joint in-store and refunded: 90/1200 = 0.075.
  3. Independence: P(refund|online)=300/750=0.40; P(refund|store)=90/450=0.20. Unequal → not independent in this sample.

Trap “Given/among” changes the denominator. “And/both” keeps the grand total. The claim concerns this sample, not customers in general.

14 · Worked 2 — StandardiseCh 2 worked

Data: both branches average $40,000 weekly sales. SDs: A $2,000; B $8,000. Both post $46,000.

A($46k−$40k)/$2k = +3.00B($46k−$40k)/$8k = +0.75
F2.2 values: equal sales; unequal standing inside each history.

Answer: A’s week is more unusual: +3 SD versus +0.75 SD. Ask A what changed and whether it can repeat; B is within normal variation and should focus on reducing variability. Raw $46,000 alone cannot rank the results.

If you seeDo
same unit, different spreadstandardise within each group
mean far above medianreport right skew; prefer median for typical case
extreme observationinvestigate; do not delete automatically
histogram changes with binstry widths; keep stable shape; state width

15 · Worked 3 — Confidence IntervalCh 3 worked

Data: n=36 invoices, x̄=$44, s=$12. Approximate 95% interval; df=35, t*=2.03.

SE12/√36=2MARGIN2.03×2=4.06INTERVAL44±4.0639.944448.06
Worked path: SE → margin → interval.
  1. SE = 12/√36 = $2.
  2. Margin = 2.03×2 = $4.06.
  3. CI = 44±4.06 = $39.94 to $48.06 (about $40–$48).
  4. Sentence: the procedure brackets the population mean and succeeds in about 95 of 100 repetitions.

Qualify: this covers the mean, not individual invoices; it is only as defensible as the sampling frame. One month or one client can make a precise answer for the wrong population.

16 · Worked 4A — Fit the LineCh 4 worked

Data: spend x=1…7 thousand; weekly sales y=3.2, 4.1, 4.4, 6.0, 6.3, 7.9, 8.1 thousand. x̄=4; ȳ≈5.71; deviation cross-products sum 24.2; squared x-deviations sum 28.

ŷ = 2.26 + 0.864xobserved x-range: 1 to 7 thousand
F4.1 values: seven stores and the least-squares line.

b₁=24.2/28=0.864; b₀=5.71−0.864×4≈2.26. Each extra $1,000 advertising is associated with about $864 more weekly sales across these stores. It does not quite pay for itself on this evidence.

Refuse x=$30,000: the data cover $1,000–$7,000; the straight relationship may not continue. Also refuse causation: budgets were observed, not assigned.

24 · Classic Trapsall chapters

Wrong moveRepair
conditional ÷ grand totaldivide inside stated row/column
large n fixes biasonly sampling noise shrinks
95% of individuals in CICI is for the population mean
r means causeassociation; name third variables

17 · Worked 4B — Test the SlopeCh 5 worked

Data: n=40 stores; fitted slope 0.86 thousand sales per thousand spent; SE=0.31.

H₀: β₁=0 · H₁: β₁≠0 · α=.05t = .86/.31≈ 2.77df=38two-sided p≈.009.009<.05 → reject H₀evidence of linear association
Worked path: population hypotheses → t → p → bounded conclusion.

There is evidence of a linear association in the population represented by these stores. Estimated size: about $860 more weekly sales per extra $1,000 spent. No causal claim; no extrapolation beyond sampled spend.

Bad sentenceRepair
“H₀ has a 0.9% chance.”p is probability of data this extreme if H₀ true.
“H₀ is proved” when p large.insufficient evidence; real small effects may remain.
“Significant = important.”judge estimate and CI in business units.
choose one-sided after resultdirection comes from question before data.

18 · Worked 4C — InteractionCh 6 worked

Model: weekly spend = 20 + 0.24h + 6d + 0.66(hd), where member d=1.

d=0: 20+0.24hd=1: 26+0.90hh=10
Interaction redraw: dummy shifts intercept; product term tilts slope.

Non-member: d=0 → ŷ=20+0.24h. Member: d=1 → ŷ=26+0.90h. Slope difference 0.90−0.24=0.66, exactly the interaction. Member at h=10: 26+9=$35.

Check If reconstructed slopes do not differ by b₃, you put the dummy coefficient into a slope. b₂ shifts the intercept; b₃ changes the slope.

The member coefficient is a comparison at h=0, not a universal member premium. With an interaction present, every main effect is conditional on the other interacting variable being zero.

If you seeWrite
coefficient changes after adding zit now compares cases alike on z; the question changed

19 · Worked 5A — ClassifierCh 9 worked

Held-back n=400: TP=30, FP=20, FN=50, TN=300; 20% churn.

PREDICTEDchurnstaychurnstay30 TP50 FN20 FP300 TNacc .825 · sens .375 · precision .60do-nothing accuracy = .80
Worked matrix: the headline barely beats the base-rate benchmark.
  1. Accuracy=(30+300)/400=0.825.
  2. Sensitivity=30/(30+50)=0.375.
  3. Precision=30/(30+20)=0.60.
  4. No-flags benchmark = non-churn share = 0.80; gain is slight.

For retention, misses are expensive; catching under two fifths of churners is weak. Lower threshold → sensitivity rises, precision falls. Cheap contact relative to lost customer makes that trade defensible.

20 · Worked 5B — Logistic ModelCh 10 worked

Model: late-payment log odds η=−4.0+0.08(days)+0.7(new), new=1.

η = −4 + .08 days + .7 new+1 DAYe^.08≈1.083NEWe^.7≈2.030 DAYSη=−1.6 → p≈.17η=0 → .08 days=4 → tipping point 50 daysuse 50 only inside observed range
Worked path: coefficient → odds multiplier → probability → tipping point.

Each extra credit day multiplies late-payment odds by about 1.083, holding customer status fixed; ten days multiplies odds by about 2.2. A new customer has about 2.0× the odds versus established on the same terms.

Established, 30 days: η=−4+2.4=−1.6 → p≈0.17. Tipping point: η=0 → days=50; beyond about 50 days, late payment becomes more likely, provided 50 is within the data’s range.

Decision cueMove
missed positive is expensivelower threshold; sensitivity ↑, false positives ↑
false action is expensiveraise threshold; precision tends to ↑, sensitivity ↓
rare positive classstate base rate before accuracy
logistic coefficient requestedreport sign, eᵇ odds multiplier, holding-fixed clause

21 · R Error DecoderCh 7 worked

OBJECTFUNCTIONARGUMENTVALUEnameverb()insideresultnot foundnot loadedunusedwrong shapelast is dangerous: no red error
F7.1 redraw: match the message to the failed hop.
Message/resultCause → fix
object not foundname not created/misspelt → run earlier line/check name
could not find functionpackage not loaded or verb misspelt
unused argumentargument name wrong
three times expected rowsnon-unique join key multiplied matches

Read the first error, not the last. Predict result rows, columns and grain before running code. Check a joined row count against the input; an increase can expose duplicate keys.

22 · Wrangling Sequence DrillCh 8 worked

RAW6×3saleFILTER4×3saleMUTATE4×4saleSUMMARISE2×2councilonly summarise changes what one row meanscarry n beside every group statistic
F8.1 redraw: rows, columns and grain are different questions.

Faulty sequence: rank 313 suburbs by median price → keep top 20 → group by council → report council means. Those are means of each council’s surviving expensive suburbs, not of its suburbs.

Repair: start with all 313 suburbs → group by council → summarise every suburb (include count) → arrange the council table. A council with one very expensive suburb no longer wins unfairly.

If you seeDo / say
top within a subsetfilter first, rank after
groups of unequal sizedo not average their averages
missing after joininvestigate; distinguish key mismatch from genuine absence
more rows after jointest key uniqueness
grouped outputstate new grain and carry count

23 · Method ChooserCh 1–10

If the question asks…Use / showGuardrail
category counts/sharesfrequency, relative frequency, barno mean of labels
two categorical variablescontingency + conditionalsdenominator follows “given”
typical skewed valuemedian + IQRmean follows tail
compare unlike spreadsz=(x−x̄)/sname reference group
population mean rangex̄±t*s/√nprocedure, not individuals
linear joint movementscatter + rr≈0 may hide curve
predict numerical yleast-squares lineresiduals + x-range
generalise sample effecthypothesis test + CIpopulation parameter
hold variables fixedmultiple regressionno omitted-variable immunity
group-specific slopesdummy + interactionb₃ = slope difference
predict a categorythreshold + confusion ratesbase rate + held-back data
binary probability modellogistic regressioncoefficient is log-odds scale
NAMEquantityFIXdenominatorCOMPUTEwith unitsLIMITclaimprobability · interval · slope · rate
F12.1 redraw: the reusable four-beat answer.

24b · Model + Data TrapsCh 5–10

Wrong moveRepair
accuracy alonebase rate + sensitivity + precision
logistic b = probability changeeᵇ multiplies odds
join silently trustedpredict and verify row count
mean of group meansuse original rows or weighted mean

25 · Exam-Morning Planstudy strategy

  1. Recite: type → display; denominator; centre/spread/shape; z; SE/CI; scatter/r/line/residual/R²; H₀/H₁/p/errors; holding-fixed/dummy/interaction; matrix/threshold; odds/log odds.
  2. Redo the five numbers cold: .325/.075; +3/+0.75; $39.94–$48.06; 0.864/2.26; .825/.375/.60; p≈.17 and 50 days.
  3. For every answer: name → denominator/scale → compute with units → limit.
https://www.asksia.ai/cheatsheet/monash-etc1000 · Full ETC1000 study guide →
v2 · 2026-10-25
Forward this to your tutorial groupShared by ________
Check your exam instructions on the LMS for what you may bring; use this sheet to revise and self-test.
Monash Semester 2 final assessment period, 2–18 November 2026 (check your personal exam timetable).

Want one for YOUR exact syllabus?

Sia is your free desktop study agent. Drop your Monash University ETC1000 slides — Sia builds a sheet tailored to YOUR exam. Better than this library because it knows YOUR materials.

↓ Download Sia · Free