PSLE Science Reality Lab Vol No.140 | “42 Samples Were Positive” — How Much Did the Cutoff Create the Count?

PSLE-SCI-REALITY-0140

Wait, What? The Samples Did Not Change, but the Number of “Positive” Samples Did

Seven samples are measured. Their results are 46, 48, 49, 50, 51, 53 and 55 units.

A report uses a rule: results of 50 units or more are labelled positive. Four samples are therefore positive.

Now imagine that another report uses a different rule: results of 52 units or more are labelled positive. Only two samples are positive.

Nothing happened to the samples. No measurement changed. The classification rule changed.

That does not mean cutoffs are dishonest or arbitrary. Many scientific and engineering decisions need thresholds. A threshold can be carefully justified by evidence, standards, method capability or the purpose of a decision. But once a continuous measurement is turned into a category such as positive/negative, pass/fail, high/low or alert/no alert, the reader should remember that the category count is partly a result of the decision rule.

The Reality Lab habit is: when a headline gives you a count of labelled cases, find the cutoff that created the labels before interpreting the count.

Quick Answer

Find the original measurements if they are available.
Locate the exact rule used to create the categories.
Check whether values close to the cutoff could switch label because of measurement uncertainty, rounding or a slightly different decision rule.
Ask why the threshold was chosen and whether it fits the intended scientific or engineering purpose.
Do not treat “positive” as a new physical substance or state unless the science really supports such a boundary.
Where the decision is important, inspect both the classification count and the underlying continuous data.

The Exact Learner Job This Page Owns

This page owns one evidence-transfer job: evaluating a scientific report, infographic, dashboard or product claim that converts continuous measurements into categories and then presents the category count as though it came directly from nature.

It does not become a general statistics page, a medical diagnosis page or a legal/regulatory guide. It does not replace the existing PSLE Science owners for measurement, variables, data tables, uncertainty, fair comparison or conclusion writing. It applies those skills to one durable communication object: the thresholded count.

Original Reality Lab Case: Seven Sensor Readings Become a Headline

This is an original composite teaching case. The values are constructed for learning and are not taken from a medical, environmental or commercial dataset.

Sample	Measured value	Label if cutoff is ≥50	Label if cutoff is ≥52
A	46	Negative	Negative
B	48	Negative	Negative
C	49	Negative	Negative
D	50	Positive	Negative
E	51	Positive	Negative
F	53	Positive	Positive
G	55	Positive	Positive

At a cutoff of 50, the public headline could say, “Four of seven samples were positive.” At a cutoff of 52, the headline becomes, “Two of seven samples were positive.”

Both statements can be mathematically correct under their own rules. Neither statement is understandable unless the reader knows the rule.

Observed, Classified and Claimed

Layer	What happened
Observed or measured	Seven numerical values were obtained: 46, 48, 49, 50, 51, 53 and 55.
Decision rule	A threshold was chosen, such as ≥50.
Classification	Each value was assigned to positive or negative.
Summary	The number of positive labels was counted.
Possible overclaim	“Nature contains exactly four fundamentally different positive samples.”

The classification is not fake. It is a decision layer built on the measurements. Scientific communication becomes misleading only when that layer is hidden or treated as though the labels appeared directly from the physical world without a rule.

A Cutoff Can Be Useful Without Being a Physical Cliff

Imagine two samples measuring 49.9 and 50.1 units under a rule where 50 is the cutoff. They receive different labels. Physically, they are very close.

This can be perfectly acceptable if the purpose requires a yes/no decision. A machine may need to trigger an alarm. A manufacturing line may need a pass/fail rule. A field programme may need to decide which samples receive follow-up analysis.

But the label difference should not make us imagine that 49.9 and 50.1 suddenly became far apart in the underlying measurement.

The Representation Check: Did the Graphic Hide the Measurements?

Suppose an infographic shows 100 dots: 42 red and 58 grey. The caption says, “42 positive samples.” The display may be useful, but it hides several things:

how far above or below the cutoff each result was;
whether many results sat just beside the threshold;
whether the cutoff changed between reports;
whether the measurement uncertainty is large compared with the distance to the threshold;
whether different methods use the same cutoff or measurement basis.

A classification plot compresses information. The learner should know what was compressed.

The Measurement-Uncertainty Check: What Happens to Values Near the Boundary?

A result reported as 50.2 units may not mean that the true quantity is known exactly as 50.2. Every real measurement has uncertainty and method limits. If the cutoff is 50 and the measurement uncertainty is large enough to cross that boundary, the classification may need a declared decision rule.

NIST and other metrology organisations treat conformity decisions carefully because the measured value, uncertainty, specification limit and decision rule are different objects. Reality Lab Vol No.126 owns that full near-limit problem. Here, the learner only needs the transfer: a count of positives can be sensitive to uncertain values near the cutoff.

The Rounding Trap: 49.6 Can Become 50 on the Screen

Suppose the actual stored result is 49.6 but a dashboard displays whole numbers. The screen shows 50.

If the classification rule is applied to the unrounded 49.6, the result may be negative. If a reader sees only the rounded 50 and assumes the rule was applied afterward, the classification can look inconsistent.

Good reports make clear whether rounding is only for display or whether the rounded value is used in the decision.

The Baseline Check: Did the Cutoff Stay the Same Over Time?

A time series can show a sudden jump in “positive cases” even if the underlying measurements changed very little when the classification threshold or method changes. The report should therefore identify changes in decision rules as well as changes in measurements.

This is related to Reality Lab Vol No.048, which asks whether a measuring method changed when a trend jumps. In the present job, the instrument can remain unchanged while the classification rule changes.

Why Scientists Use Decision Thresholds

Thresholds can be based on many things: method performance, reference distributions, engineering requirements, risk-management choices, specification limits or a practical need to trigger further action. The exact basis depends on the field.

A Primary learner should not invent a universal threshold rule. Instead, ask three questions:

What quantity was measured?
What cutoff was applied?
Why is that cutoff suitable for the purpose?

The scientific strength comes from making the rule explicit and testing whether it matches the intended decision.

What Evidence Would Strengthen “42 Samples Were Positive”?

The measurement method and units are stated.
The exact cutoff and whether equality counts are stated.
The scientific or engineering basis for the threshold is explained or linked.
Individual measurements or their distribution are available, especially near the cutoff.
Measurement uncertainty and rounding are handled consistently.
The same classification rule is used for comparisons across groups or times—or changes are explicitly documented.
The report distinguishes screening labels from final conclusions when further testing is required.

What Would Weaken It?

The report gives a positive count but hides the threshold.
The cutoff changes without being disclosed.
Most results sit extremely close to the boundary but are shown as sharply different groups.
Rounded display values are mistaken for the values used in classification.
Two studies compare positive counts generated by different methods or thresholds.
A screening classification is described as a physical certainty.

Worked Case 1: The Quality-Control Line

A factory measures a property of five parts: 9.7, 9.9, 10.0, 10.1 and 10.3 units. A rule labels values of 10.0 or above as “high”. Three are high. The useful conclusion is about the rule-based classification. It is not that the five parts naturally divide into two distant types.

Worked Case 2: The Sensor Alert

A field sensor triggers an alert above 80 units. A reading of 79.8 does not trigger; 80.2 does. The alert system needs a boundary to operate. But if the sensor uncertainty is ±2 units, the two physical conditions may not be meaningfully distinguishable from one reading alone. The decision rule and measurement uncertainty should both be visible.

Worked Case 3: Two Reports, Two Cutoffs

Report A defines “high” as ≥40 and says 60 of 100 samples are high. Report B defines “high” as ≥50 and says 35 of 100 are high. It is invalid to conclude that the second population must have fewer high measurements without first applying a comparable definition or examining the underlying values.

Worked Case 4: The Hidden Distribution

Two datasets each contain 50% positive samples at a cutoff of 50. Dataset X clusters tightly around 49–51. Dataset Y has half the values near 10 and half near 90. The same positive percentage hides very different evidence patterns.

Tempting Reasoning That Fails

“Positive means a completely different physical state.” Sometimes it is simply a decision label above a threshold.
“A cutoff is arbitrary, so the classification is useless.” A cutoff can be carefully justified and very useful.
“The count is objective, so the rule does not matter.” The count depends on the measurements and the classification rule.
“Anything just below the threshold is safe or absent.” The meaning depends on the quantity, purpose and uncertainty; avoid importing legal or health conclusions.
“Changing the threshold proves somebody manipulated the data.” Standards and scientific understanding can change legitimately; the important requirement is transparency and comparability.
“If 42% are positive, the average must be 42% of the cutoff.” A category percentage does not determine the average measurement.

Model and Measurement Limits

This page deliberately does not teach diagnostic testing, clinical sensitivity/specificity, legal limits or regulatory compliance. Those domains require specialised authority and context. The transferable PSLE Science job is simpler: understand how a numerical measurement becomes a category and keep the category’s meaning attached to its rule.

A threshold also cannot repair a weak measurement. If the instrument cannot reliably distinguish values around the cutoff, a clean-looking positive/negative label can create more certainty than the evidence supports.

How Far Can the Conclusion Travel?

If a clear, justified threshold is applied consistently to valid measurements, a positive count can accurately describe how many results met that rule. It does not automatically tell us the average magnitude, the full distribution, how far each value lies from the boundary, whether another threshold would produce the same ranking, or whether the categories represent permanent natural kinds.

PSLE-Style Transfer Case

A sensor measures five samples: 18, 19, 20, 21 and 22 units. A project labels results of 20 or more as “high”.

Question: A pupil says, “There are three high samples, so the three must be very different from the other two.” Why is this not justified?

Reasoned answer: The measurements form a gradual sequence. The high/low distinction comes from a cutoff at 20. Values of 19 and 20 are close even though they receive different labels. The classification is useful for the stated rule but does not prove a large physical gap.

Explained Practice

Practice A: Measurements are 49, 50 and 51; cutoff ≥50. How many are positive? Two. What else should a careful reader notice? All three measurements are close to the decision boundary.

Practice B: A headline says “positives doubled” after the threshold changed from 60 to 50. Can the count change without any physical measurement changing? Yes. The classification rule itself can change the count.

Practice C: Two studies use cutoffs 40 and 55. Can their positive percentages be compared directly? Not without understanding the different definitions and underlying measurements.

Delayed Independent Return: The C-U-T-O-F-F Check

C — Continuous value: What was actually measured?
U — Units: Are the quantity and units clear?
T — Threshold: What exact cutoff creates the label?
O — Origin: Why was this threshold chosen?
F — Fringe: How many results sit close to the cutoff?
F — Fair comparison: Did all groups and times use the same rule?

Parent and Tutor Teaching Guide

Place number cards 46, 48, 49, 50, 51, 53 and 55 on a table. Put a string between 49 and 50. Ask the learner to count the cards on the “positive” side. Then move the string between 51 and 53 without moving any number card. The count changes immediately.

Ask the learner what changed: the samples, the measurements or the rule? Then ask a deeper question: can the new rule still be justified? This prevents the lesson from turning into cynical “all thresholds are arbitrary” thinking. Scientific maturity means recognising that rules can be useful while still understanding how they shape summaries.

Authoritative Sources

The current Singapore Science framework asks learners to interpret and analyse information, evaluate methods, consider uncertainty and communicate reasoning. Turning measurements into categories is a good test of those habits because the labels can appear simpler and more certain than the evidence underneath them.

The Quiet Return

A cutoff can make a decision clearer.

It should not make the measurement underneath it disappear.

When science gives you a count of “positive” cases, find the line that created the category—and look at the numbers standing beside it.

Small Group Tutorials