DAO1704 Decision Analytics using Spreadsheets and AI
The Complete Exam Bible · Sem 1 AY26/27DAO1704 Overview
- National University of Singapore
- Semester 1, AY2026/27
- Level 1
- Closed-book final exam
- One double-sided A4 sheet
What this course asks you to do
DAO1704 uses spreadsheets as a decision environment rather than as a calculator with a large screen.
- Assessment split Class discussion, a computer-based midterm and a pen-and-paper final contribute to the published grade.
- Hardest transition Move from a spreadsheet result to a sentence that names its denominator, reference population and sensitivity.
- Revision priority Rebuild formulas and sampling relationships from memory before compressing them onto the permitted sheet.
What you get in the full DAO1704 Exam Bible
- 6 chapters, 29 pages — every examinable topic, week by week
- 19 original practice questions, each with a model answer in the Final Exam format
- 6 worked examples, solved the way DAO1704 marks
- 6 computed figures, redrawn from the course’s own numbers
- Exam-morning plan for the Final Exam (50%) — what to revise last, where marks are lost
Free on this page: assessment table, chapter map, one worked example, key terms, FAQ. The rest is in the full exam bible.
How DAO1704 is assessed
| Component | Weight | Format |
|---|---|---|
| Class Discussion | 20% | Attendance-adjusted contribution to class discussion, including preparation, thoughtful comments and respect for other views. |
| Midterm Test | 30% | Weeks 2 to 6. Two-hour, computer-based, closed-book test. Microsoft Excel, one double-sided A4 cheat sheet and a calculator are permitted. |
| Final Exam | 50% | Weeks 6 to 13. Two-hour, pen-and-paper, closed-book examination. One double-sided A4 cheat sheet and a calculator are permitted. |
The three published numeric weights total 100%. Check the current Canvas task pages for any component-specific pass requirements or operational changes.
What other students are asking about DAO1704
- Where can I find lecture notes or study guides for DAO1704 at NUS?
- The DAO1704 study guide in AskSia Library covers Decision Analytics using Spreadsheets and AI in 6 chapters, with the assessment table, a chapter-by-chapter topic map, key terms and a worked example, free to read on this page.
- Is there a final exam in DAO1704, and how much is it worth?
- Yes. In the offering this AskSia Library guide is built from, Final Exam carries 50% of the DAO1704 mark; the full split is Class Discussion 20%, Midterm Test 30%, Final Exam 50%.
- What does DAO1704 cover on visual choices and pivottable evidence?
- In AskSia Library's DAO1704 guide, the Visual Choices and PivotTable Evidence chapter explains match variable types to comparisons and challenge every aggregation.
- What does DAO1704 cover on repeatable data preparation with power query?
- In AskSia Library's DAO1704 guide, the Repeatable Data Preparation with Power Query chapter explains record data transformations, preserve grain and verify every refresh.
- What does DAO1704 cover on descriptive measures and distribution traps?
- In AskSia Library's DAO1704 guide, the Descriptive Measures and Distribution Traps chapter explains choose centre and spread from shape, purpose and sensitivity.
Current DAO1704 dates
| Date | Item | Control |
|---|---|---|
| 3 October 2026 | Midterm Test | The current Canvas announcement gives the test time as 12:00 pm to 2:00 pm. |
| 21 November 2026 | Final Exam | The course administrative deck gives the start time as 9:00 am. |
Dates are as published in the Semester 1, AY2026/27 course materials. Confirm exact deadlines and submission settings in the live LMS.
What DAO1704 covers
Six source-led chapters move from auditable formulas and visual evidence through replayable data preparation, distribution diagnosis, grouped comparisons and repeated-sample inference.
Spreadsheet Logic and Auditable Formulas
Control copied references, conditional boundaries and lookup assumptionsFormula structure before output
A spreadsheet formula is a reusable decision rule. Relative references move with the formula, absolute references remain anchored, and mixed references lock only one dimension.
- Relative, mixed and absolute cell references
- Visible assumptions in control cells
- IF and ordered IFS boundaries
- Aligned criteria ranges in conditional functions
- XLOOKUP keys and missing matches
- Copy-down and perturbation testing
Visual Choices and PivotTable Evidence
Match variable types to comparisons and challenge every aggregationBegin with variable type
A visual should make the intended comparison easier to see.
- Chart choice by variable type
- Histogram width and binning sensitivity
- Boxplots, quartiles and flagged outliers
- PivotTable row, column and value roles
- Counts, sums and averages as different questions
- Scatterplots and third-variable encodings
Repeatable Data Preparation with Power Query
Record data transformations, preserve grain and verify every refreshPreparation becomes a recorded process
Power Query connects to source files and records transformations as Applied Steps. That makes a monthly workflow replayable, but it also repeats every assumption.
- Get and Transform source connections
- Applied Steps as an audit trail
- Data types and conversion errors
- Duplicate records and declared grain
- Merge keys, expansion and row multiplication
- Refresh reconciliation and control totals
Descriptive Measures and Distribution Traps
Choose centre and spread from shape, purpose and sensitivityOne average is not a distribution
The mean uses every value and responds to extremes, while the median follows order and resists them.
- Mean, median and mode as centres
- Range, MAD, variance and standard deviation
- Quartiles and interquartile range
- Boxplot fences and source validation
- Skew and robust summaries
- Sensitivity statements for decisions
Grouped Comparisons and Simpson's Paradox
Separate performance from case mix with stratification and standardisationOverall rates carry the observed mix
An aggregate rate is a weighted average of rates inside subgroups. When compared groups face different proportions of easy and hard cases, the overall comparison combines performance with allocation.
- Aggregate rates as weighted averages
- Case mix and denominator visibility
- Confounder conditions
- Stratified like-with-like comparisons
- Standardisation to a common mix
- Operational and fair-comparison conclusions
Sampling Distributions and Estimator Precision
Connect probability samples to unbiasedness, variance and standard errorStatistics vary because samples vary
A probability sample connects a subset to a population through a known selection mechanism. Draw another valid sample and the sample mean usually changes.
- Simple random sampling without replacement
- Statistics as sample-dependent random variables
- Sampling distributions across repeated samples
- Expected value and unbiasedness
- Variance and estimator efficiency
- Standard error and square-root returns
The captured teaching sequence begins with formula logic and data visualisation, moves through replayable preparation and descriptive analysis, and reaches sampling and estimation before the later decision-modelling block. At every stage, the output is only the middle of the answer.
The useful work is identifying what the number measures, which assumptions created it and whether a nearby defensible choice would change the decision.
Why spreadsheet control matters
Copied references, criteria ranges and lookup keys can generate thousands of plausible answers from one quiet mistake. The first chapter therefore treats a formula as a rule with moving and fixed parts.
It shows how to expose global assumptions in control cells, lock them deliberately, test exact boundaries and check range alignment. This method turns a workbook from an opaque result into a traceable model.
How visual evidence can mislead
Charts inherit every aggregation decision beneath them. A histogram can reveal or suppress shape through its bin width.
A PivotTable can count a field when the intended summary was a sum. A regional total can lead because of more transactions even while the average transaction is lower.
The guide keeps variable type, aggregation and denominator visible so that a chart answers the stated question instead of a more convenient one.
From manual cleaning to replayable preparation
Power Query records data preparation as an ordered process: connect, inspect, transform, merge, load and refresh. That order creates a review trail, but it does not make the assumptions self-validating.
Incoming files can change schema, a date can import as text, duplicate keys can multiply rows and a refresh can replay an old error perfectly. The preparation chapter uses row counts, grain definitions and merge checks as the receipt for a trustworthy refresh.
From description to inference
Mean, median and spread answer different questions about a distribution. Grouped rates can change when case mix changes.
A sample statistic changes because another valid sample could have been drawn. The final chapters connect these ideas: diagnose shape before choosing a summary, inspect meaningful strata before comparing groups, and distinguish unbiasedness from precision before judging an estimator.
The worked examples use original data and show every arithmetic step.
How the assessment changes revision
The published midterm is computer based, covers Weeks 2 to 6 and permits Microsoft Excel, a calculator and one double-sided A4 sheet. The published final is a two-hour, pen-and-paper closed-book exam covering Weeks 6 to 13 with the same paper-sheet and calculator allowances.
Sampling and estimation sits at the overlap. Revision should therefore include both workbook fluency and manual setup: calculate in Excel, then reconstruct the relationship and interpretation without it.
Worked example · free
Building a sampling distribution from every possible sample
- 2List the possible samples and their means.
- 2Use the equal probabilities to find the expected sample mean.
- 2Calculate the probability-weighted squared deviations.
- 2Interpret unbiasedness and remaining variability separately.
Key terms
- Absolute Reference
- A cell reference whose locked row, column or both remain fixed when a formula is copied.
- Criteria Range
- The cells tested by a conditional function such as COUNTIFS or AVERAGEIFS.
- PivotTable
- An Excel summary that assigns fields to rows, columns, filters and aggregated values.
- Binning Bias
- A change in apparent histogram shape caused by the chosen bin width or boundaries.
- Data Grain
- The real-world unit represented by one row of a table.
- Applied Step
- A recorded Power Query transformation replayed in sequence during refresh.
- Interquartile Range
- The distance from the first quartile to the third, describing the middle half of observations.
- Confounder
- A third variable linked to the outcome and distributed unevenly across compared groups.
- Standardisation
- Recombining stratum-specific rates under one common reference mix.
- Sampling Distribution
- The probability distribution of a statistic across all repeated samples under a sampling design.
- Unbiased Estimator
- An estimator whose expected value equals the population parameter it targets.
- Standard Error
- The standard deviation of an estimator across repeated samples.
DAO1704 FAQ
How is the published grade divided?
Class Discussion contributes 20%, the Midterm Test contributes 30%, and the Final Exam contributes 50%. The numeric weights total 100%. The discussion score is adjusted by attendance, while the two tests have distinct delivery formats and coverage periods.
What is the midterm format?
The current Canvas announcement describes a two-hour, computer-based test covering Weeks 2 to 6. Internet access is blocked, but Microsoft Excel, one double-sided A4 cheat sheet and a calculator are permitted. The announcement identifies multiple-choice, fill-in-the-blank and open-ended questions.
What is the final exam format?
The course administrative deck describes a two-hour, on-campus, pen-and-paper closed-book examination covering Weeks 6 to 13. One A4 cheat sheet written on both sides and calculators, including graphing calculators, are permitted.
How should a copied spreadsheet formula be checked?
Classify every reference as row-specific, column-specific or global, then apply relative, mixed or absolute locking accordingly. Test an exact boundary and values on both sides, inspect the range endpoints, and change one shared assumption to confirm that only the intended outputs move.
What makes a Power Query refresh trustworthy?
A trustworthy refresh preserves the declared row grain, assigns correct types, reconciles row counts around destructive steps, identifies conversion errors, checks unmatched and duplicate merge keys, and agrees with a stable control total. Successful completion alone does not prove the incoming schema still matches the recorded process.
Why can an overall group comparison be misleading?
An overall rate is a weighted average of stratum rates. If compared groups face different mixes of easy and hard cases, the weights combine performance with allocation. Show the within-stratum rates and denominators, then standardise to a stated common mix when a fair process comparison is needed.
What does a sampling distribution add to one sample?
It represents how a statistic would vary across all samples generated by the design. Its expected value addresses long-run centring and its variance addresses precision. One observed sample supplies a statistic, but it cannot display those repeated-sampling properties by itself.
Questions students ask in each DAO1704 chapter
- Why can a copied formula be wrong when the first result is right?
- The first row can use the intended assumption by accident. Once copied, an unlocked global reference moves to a different cell and the error begins. Inspect the reference pattern and test the next row rather than validating only the original value. Chapter 01 · Spreadsheet Logic and Auditable Formulas
- How should a histogram bin width be chosen?
- Try several nearby widths and retain conclusions that persist. Too few bins can hide clusters or skew, while too many can make random fluctuation look structural. Cross-check with quartiles and a boxplot before describing the distribution. Chapter 02 · Visual Choices and PivotTable Evidence
- Why is a completed refresh not enough evidence?
- The sequence can finish while the source schema, units or key uniqueness have changed. Reconcile row counts, errors, unmatched keys and a control total so that successful execution is paired with evidence that the result still represents the intended table. Chapter 03 · Repeatable Data Preparation with Power Query
- When should the median be preferred to the mean?
- Prefer the median when the decision concerns a typical observation and strong skew or extreme values pull the mean away from most cases. Keep the mean when totals or additive planning matter, and explain the difference rather than replacing one silently. Chapter 04 · Descriptive Measures and Distribution Traps
- Can case mix matter without a full reversal?
- Yes. Unequal workloads can compress or exaggerate an overall difference even when the winner remains the same. Show the subgroup rates and apply a common mix when the size of the process difference affects the decision. Chapter 05 · Grouped Comparisons and Simpson's Paradox
- Why is a statistic treated as random?
- Its value depends on which units enter the sample. Before selection, several valid samples and corresponding statistics are possible. The sampling distribution represents that design-based uncertainty even though the population parameter itself is fixed. Chapter 06 · Sampling Distributions and Estimator Precision
How to study for the exam
Keep one small workbook and one handwritten sheet throughout revision. In the workbook, rebuild each chapter's method with fresh values: copy a formula, change a bin width, refresh a query, compare robust summaries, standardise a grouped rate and enumerate a sampling distribution. On paper, write the relationship, denominator, assumption and direction check without looking.
When the two disagree, diagnose the reference or interpretation before repeating the calculation. Because the midterm permits Excel and the final is pen and paper, practise moving both ways: from a formula to plain language and from a decision sentence back to a formula. Finish every session with one original practice question whose answer includes a calculation and a limitation. The purpose is not to collect more examples.
It is to make the audit route automatic under time pressure. In the final week, compress only the relationships you have repeatedly failed to retrieve onto the permitted sheet, and keep units and applicability conditions beside each one.
Your AI Statistics tutor for DAO1704
Stuck on a hard DAO1704 question? Sia is AskSia’s AI Statistics tutor — ask any DAO1704 Decision Analytics using Spreadsheets and AI question and get a clear, step-by-step explanation grounded in how the course is actually taught and assessed. Read this whole study guide free, then take your hardest questions to Sia.
