CDT502 Chap.1 From Data to Decisions
From Data to Decisions
This chapter separates observed records, descriptive summaries, population inference and operational action. It shows why a correct calculation is only one rung in an evidence chain and why every percentage needs a denominator, every generalisation needs a sampling route, and every recommendation needs an objective and a stated cost of error.
The worked service-desk example compares mean and median without treating robustness as permission to ignore an extreme day. Start by identifying the unit of observation and the process that produced each row. A customer, transaction, branch and day can all appear in one file, yet they answer different questions and require different denominators.
Before calculating, rewrite the managerial prompt as a claim about one unit, outcome, period and choice. This exposes requests that look numerical but actually need a definition or data-collection decision first. Next separate a statement about recorded cases from a statement about a population. The first can be checked against the file.
The second needs a credible route beyond it, such as a sampling design or stable operational coverage. Adding a recommendation creates another layer because the action depends on objectives, constraints and the relative cost of errors. Keep these layers in separate sentences so a reader can challenge the exact assumption that carries the conclusion. Directional prediction is a useful check.
If one high observation is added, anticipate how mean, median and range should respond. If a denominator expands while the numerator stays fixed, predict the rate movement before dividing. These simple forecasts catch transcription mistakes and force the statistic to retain its meaning.
The final habit is to end with a bounded decision: state what can be done now, what remains uncertain and which next observation would change the choice. That closing move turns analysis into accountable support rather than numerical decoration.
What this chapter covers
- 01
Data, information and decision evidence
- 02
Descriptive versus inferential claims
- 03
Units, denominators and scope
- 04
Mean and median under skew
- 05
Uncertainty before action
- 06
Bounded operational recommendations
Worked example · free
Separate a typical day from a capacity decision
- 2Calculate the mean from all eight observations.
- 2Order the values and calculate the median.
- 3Recommend an action that respects the evidence scope.
Key terms
- Evidence Scope
- The population, period and conditions to which a result can defensibly apply.
- Descriptive Statistic
- A numerical summary of values that were observed.
- Inferential Claim
- A claim extending sample evidence to a wider population under a design and assumptions.
From Data to Decisions FAQ
Why is a correct average not enough for a decision?
An average compresses observed values but does not show spread, selection, unusual cases or the cost of acting incorrectly. A decision also needs the evidence scope, uncertainty and objective. Report the average as one input, then state which action it supports and which additional evidence would change that action.
What is the quickest way to stop overclaiming?
Write the unit of observation, denominator and period before interpreting the number. Then phrase the descriptive result using the observed dataset rather than a population noun. Add the wider population only after identifying a credible sampling or data-collection route from those observations to that population.
How should I respond to one extreme observation?
Verify its value, unit and process context first. Compare summaries with and without it as a sensitivity analysis, but do not delete it solely because it changes the mean. A genuine extreme may reveal the operational failure the decision must address, even when a median better describes ordinary cases.
Exam move
Rebuild the evidence ladder from memory: records, summary, uncertainty and decision. Take any dashboard statement and rewrite it four times, once at each rung. For each worked problem, circle the denominator and underline the population noun. If the data-collection route does not support that noun, narrow the claim before doing more arithmetic. Create a two-column error log.
In the first column, copy only the claim you originally made; in the second, identify whether the defect came from scope, summary choice, uncertainty or action. Rework the same prompt with a different observation period or decision cost and note which rung changes. Once a week, choose a news graphic or workplace metric and reconstruct the missing unit, denominator and collection path.
Finish by stating one reversible step that the current evidence supports and one irreversible step it cannot yet support. This practice makes the boundary visible before examination wording changes the surface details.
Working through From Data to Decisions in CDT502? Sia is AskSia’s AI Statistics tutor — ask any CDT502 From Data to Decisions question and get a clear, step-by-step explanation grounded in how CDT502 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
