Why Mathematics? | Greenwood Variance for Reliability Curves: Risk Sets, Kaplan–Meier Uncertainty, Confidence Intervals and Tail Instability

A single Kaplan–Meier step curve hides sampling variation from finite event counts and censoring. Two small fleets can show visibly different curves by chance. This is the central mathematical mechanism in why a curve needs uncertainty. Write the quantities and assumptions before using software, then reproduce the calculation or geometric argument in a small example. A narrow line on a graph is not evidence of precision. Add intervals and risk-set counts. The useful habit is to keep the physical question, the probability model and the numerical method in separate columns so that a correct computation is not mistaken for complete evidence.

Did You Know? A result can be numerically stable and still answer the wrong question if the failure event, denominator or transformation changes. Test one nearby case and predict the direction of change before recalculating. Save inputs, intermediate values and units so a second reader can audit the path.

For practice, turn the example into a sensitivity check. Change one defensible input while holding the others fixed, compare the new result with the original and explain why it moved. Then construct a counterexample in which the shortcut fails. End with two sentences: what the mathematics supports, and what remains uncertain.

Contents · Next chapter

Start with the operational question behind kaplan–meier foundation. At event time t_j, the survival multiplier is 1-d_j/n_j, and the product of multipliers estimates S(t). In the illustrative case, One failure among 20 at risk multiplies survival by .95. Draw the relevant graph, density, boundary or timeline before calculating; the picture should use the same variables and sign convention as the equations. Independent-censoring assumptions remain essential. Build the risk table first.

A student should be able to rebuild the result without hidden settings. Record the sampling rule or optimisation criterion, the exact failure definition, and the evidence used for every distribution or parameter. Compare a transparent baseline with the advanced method. If they disagree, investigate the mechanism before deciding which answer is more credible.

This chapter belongs to a larger decision, so translate the number carefully. State one action the evidence permits and one action it does not. A parent or educator can ask which school ideas are doing the work: algebra organises relationships, probability represents uncertainty, calculus describes local change, and computing repeats the documented rules at scale.

Contents · Previous chapter · Next chapter

Treat greenwood variance sum as a model-checking problem rather than a vocabulary exercise. The estimated variance of log S-hat uses the accumulated sum d_j/[n_j(n_j-d_j)]. One failure among 20 contributes 1/(20·19). Recompute the example from the raw values and label every unit. The denominator fails when every remaining unit fails together. Inspect each event-time term.

The strongest check uses evidence the model did not choose for itself. Hold out later observations, an independent simulation set or a known-answer benchmark. Report errors where decisions are sensitive, especially near the failure boundary, instead of relying on one global accuracy score. A method that performs well on easy safe cases can still fail at the rare cases that motivated the analysis.

Now change one assumption and make a directional prediction before running the calculation. If the result moves the opposite way, inspect the sign convention, transformation, density ratio or constraint. This prediction-first routine develops transferable problem-solving skills because it connects symbolic work to a testable expectation.

Contents · Previous chapter · Next chapter

The mathematics of variance on the survival scale becomes useful only when its mechanism is explicit. By the delta method, Var[S-hat(t)] is approximately S-hat(t)^2 times the Greenwood sum. Consider the worked illustration: If S-hat=.8 and the sum=.02, variance is .0128. Preserve the original inputs, calculate step by step and compare with a second method whenever an analytic or high-budget reference exists. This is a large-sample approximation. Report the standard error as well.

There are three uncertainties to keep apart. Sampling or numerical uncertainty comes from finite computation. Parameter uncertainty comes from estimating inputs. Model-form uncertainty comes from simplified physics, distributions or dependence. Reducing the first does not automatically reduce the other two. A credible report names all three and shows which one drives the decision.

For a practical investigation, create a small reproducible notebook or spreadsheet with a frozen data sheet, formula sheet, diagnostics sheet and decision note. Another student should be able to change one input and obtain the same update without guessing what the model meant.

Primary source: https://doi.org/10.1080/01621459.1958.10501452. Read it for the method and its assumptions; do not transfer its performance claims automatically to another model.

Contents · Previous chapter · Next chapter

Use a worked calculation to practise sensitivity analysis. With S-hat=.75 and Greenwood sum .04, SE=.75·sqrt(.04)=.15. A plain 95% interval would use .75±1.96(.15). Before accepting the result, identify the quantity that would change the conclusion most. The plain interval crosses natural bounds. Prefer a justified transformation.

A simple baseline is essential. It reveals whether the advanced method is improving accuracy or only adding complexity. Compare model calls, wall time, uncertainty and failure-mode coverage under one declared budget. Repeat stochastic calculations with several seeds; a favourable single run is not evidence of dependable efficiency.

Students can extend the example by adding correlation, a second failure mode or a modest measurement error. Explain which part of the formula changes and which remains fixed. This exercise shows how mathematics transfers from a clean textbook case to engineering and computing without pretending that the classroom model is a certified design tool.

Contents · Previous chapter · Next chapter

Build a small counterexample for why contributions accumulate. Every earlier event affects the product-limit estimate at later times, so its uncertainty carries forward. The initial example is: An early small risk set can influence all later estimates. Then create a nearby case where the usual shortcut becomes unreliable. Later censoring without events changes future denominators rather than adding a direct term. Trace event times in order.

Counterexamples are positive learning tools. They reveal the conditions that make a method work and help students recognise when a software output deserves a pause. Keep the counterexample small enough to plot or enumerate. Check sign, support, bounds and conservation of probability before adding complexity.

When writing the conclusion, separate a documented fact from an interpretation and from an illustrative scenario. Cite the primary method paper for the factual mechanism, use the local calculation for the illustration, and label any recommendation as conditional on the stated model. This keeps career and engineering advice age-appropriate and honest.

Contents · Previous chapter · Next chapter

Connect risk-set size controls precision to a real decision while protecting the boundary between calculation and judgement. For one event, d/[n(n-d)] grows as n becomes small. One failure among 5 contributes far more than one among 100. First verify the arithmetic, then translate it into an allowed action and an action the evidence does not support. Late curve steps can look dramatic because of tiny denominators. Show numbers at risk.

The audit should include the original question, input provenance, failure convention, algorithm settings, convergence or stopping rule, and a sensitivity result. For an adaptive method, also preserve the sequence of choices. For an optimisation, preserve every constraint. For Bayesian updating, preserve the prior and observation model. These records turn a result into something another person can challenge and improve.

A final transfer question is useful: where else would the same mathematical idea appear? Density ratios appear in survey correction and machine learning, probabilistic constraints in planning, and conditional probability in medical testing. The context changes, but transparent assumptions and reproducible checks remain valuable.

Contents · Previous chapter · Next chapter

Use d_j greater than one at a shared event time in both the survival multiplier and variance term. Three failures among 30 yield 3/[30·27]. This is the central mathematical mechanism in handling tied failures. Write the quantities and assumptions before using software, then reproduce the calculation or geometric argument in a small example. Rounding times can create artificial ties. Preserve time resolution. The useful habit is to keep the physical question, the probability model and the numerical method in separate columns so that a correct computation is not mistaken for complete evidence.

Did You Know? A result can be numerically stable and still answer the wrong question if the failure event, denominator or transformation changes. Test one nearby case and predict the direction of change before recalculating. Save inputs, intermediate values and units so a second reader can audit the path.

For practice, turn the example into a sensitivity check. Change one defensible input while holding the others fixed, compare the new result with the original and explain why it moved. Then construct a counterexample in which the shortcut fails. End with two sentences: what the mathematics supports, and what remains uncertain.

Contents · Previous chapter · Next chapter

Start with the operational question behind censoring affects later variance. A censored unit leaves subsequent risk sets, making later event contributions larger. In the illustrative case, Heavy early censoring widens late intervals. Draw the relevant graph, density, boundary or timeline before calculating; the picture should use the same variables and sign convention as the equations. Censoring itself is not counted as failure. Mark censoring and report its pattern.

A student should be able to rebuild the result without hidden settings. Record the sampling rule or optimisation criterion, the exact failure definition, and the evidence used for every distribution or parameter. Compare a transparent baseline with the advanced method. If they disagree, investigate the mechanism before deciding which answer is more credible.

This chapter belongs to a larger decision, so translate the number carefully. State one action the evidence permits and one action it does not. A parent or educator can ask which school ideas are doing the work: algebra organises relationships, probability represents uncertainty, calculus describes local change, and computing repeats the documented rules at scale.

Contents · Previous chapter · Next chapter

Treat plain normal intervals as a model-checking problem rather than a vocabulary exercise. S-hat±z·SE is simple but can extend below zero or above one and behave poorly near boundaries. A .95 estimate with SE .06 gives an upper limit above one. Recompute the example from the raw values and label every unit. Clipping changes coverage. Use transformed intervals.

The strongest check uses evidence the model did not choose for itself. Hold out later observations, an independent simulation set or a known-answer benchmark. Report errors where decisions are sensitive, especially near the failure boundary, instead of relying on one global accuracy score. A method that performs well on easy safe cases can still fail at the rare cases that motivated the analysis.

Now change one assumption and make a directional prediction before running the calculation. If the result moves the opposite way, inspect the sign convention, transformation, density ratio or constraint. This prediction-first routine develops transferable problem-solving skills because it connects symbolic work to a testable expectation.

Contents · Previous chapter · Next chapter

The mathematics of log transformation becomes useful only when its mechanism is explicit. An interval for log S keeps back-transformed survival positive but can still behave asymmetrically. Consider the worked illustration: Exponentiating log-scale limits creates multiplicative bounds. Preserve the original inputs, calculate step by step and compare with a second method whenever an analytic or high-budget reference exists. It is awkward when S-hat reaches zero. State the transformation.

There are three uncertainties to keep apart. Sampling or numerical uncertainty comes from finite computation. Parameter uncertainty comes from estimating inputs. Model-form uncertainty comes from simplified physics, distributions or dependence. Reducing the first does not automatically reduce the other two. A credible report names all three and shows which one drives the decision.

For a practical investigation, create a small reproducible notebook or spreadsheet with a frozen data sheet, formula sheet, diagnostics sheet and decision note. Another student should be able to change one input and obtain the same update without guessing what the model meant.

Contents · Previous chapter · Next chapter

Use log-minus-log transformation to practise sensitivity analysis. Applying log[-log S] often gives bounded, asymmetric survival intervals with better behaviour near zero and one. Back-transformation reverses limit order because the map is decreasing. Before accepting the result, identify the quantity that would change the conclusion most. Implementation mistakes can swap bounds. Verify monotonicity numerically.

A simple baseline is essential. It reveals whether the advanced method is improving accuracy or only adding complexity. Compare model calls, wall time, uncertainty and failure-mode coverage under one declared budget. Repeat stochastic calculations with several seeds; a favourable single run is not evidence of dependable efficiency.

Students can extend the example by adding correlation, a second failure mode or a modest measurement error. Explain which part of the formula changes and which remains fixed. This exercise shows how mathematics transfers from a clean textbook case to engineering and computing without pretending that the classroom model is a certified design tool.

A second primary source is https://wellcomecollection.org/works/dyw3nj4k. Compare the problem definition, estimator and validation rather than quoting one result in isolation.

Contents · Previous chapter · Next chapter

Build a small counterexample for logit and arcsine alternatives. Other transformations trade boundary behaviour and small-sample properties. The initial example is: A logit interval stays within zero and one. Then create a nearby case where the usual shortcut becomes unreliable. No transformation cures a nearly empty risk set. Compare justified choices.

Counterexamples are positive learning tools. They reveal the conditions that make a method work and help students recognise when a software output deserves a pause. Keep the counterexample small enough to plot or enumerate. Check sign, support, bounds and conservation of probability before adding complexity.

When writing the conclusion, separate a documented fact from an interpretation and from an illustrative scenario. Cite the primary method paper for the factual mechanism, use the local calculation for the illustration, and label any recommendation as conditional on the stated model. This keeps career and engineering advice age-appropriate and honest.

Contents · Previous chapter · Next chapter

Connect pointwise coverage to a real decision while protecting the boundary between calculation and judgement. A 95% Greenwood interval targets coverage at one chosen time t, not the entire curve simultaneously. Intervals at 1000 hours can be valid pointwise even if some interval misses elsewhere. First verify the arithmetic, then translate it into an allowed action and an action the evidence does not support. Scanning many times raises familywise error. Use simultaneous bands for whole-curve claims.

The audit should include the original question, input provenance, failure convention, algorithm settings, convergence or stopping rule, and a sensitivity result. For an adaptive method, also preserve the sequence of choices. For an optimisation, preserve every constraint. For Bayesian updating, preserve the prior and observation model. These records turn a result into something another person can challenge and improve.

A final transfer question is useful: where else would the same mathematical idea appear? Density ratios appear in survey correction and machine learning, probabilistic constraints in planning, and conditional probability in medical testing. The context changes, but transparent assumptions and reproducible checks remain valuable.

Contents · Previous chapter · Next chapter

When few systems remain, Greenwood terms and survival steps become highly variable. A final failure among two units adds 1/[2·1]=.5 to the variance sum. This is the central mathematical mechanism in tail instability. Write the quantities and assumptions before using software, then reproduce the calculation or geometric argument in a small example. A plotted tail can overstate knowledge. Stop interpretation at a supported risk threshold. The useful habit is to keep the physical question, the probability model and the numerical method in separate columns so that a correct computation is not mistaken for complete evidence.

Did You Know? A result can be numerically stable and still answer the wrong question if the failure event, denominator or transformation changes. Test one nearby case and predict the direction of change before recalculating. Save inputs, intermediate values and units so a second reader can audit the path.

For practice, turn the example into a sensitivity check. Change one defensible input while holding the others fixed, compare the new result with the original and explain why it moved. Then construct a counterexample in which the shortcut fails. End with two sentences: what the mathematics supports, and what remains uncertain.

Contents · Previous chapter · Next chapter

Start with the operational question behind median reliability uncertainty. Confidence limits for a survival percentile can be obtained by inverting a curve interval or test. In the illustrative case, Find where a survival band crosses .5. Draw the relevant graph, density, boundary or timeline before calculating; the picture should use the same variables and sign convention as the equations. A flat or unobserved crossing can give wide or open limits. Report when the median is not estimable.

A student should be able to rebuild the result without hidden settings. Record the sampling rule or optimisation criterion, the exact failure definition, and the evidence used for every distribution or parameter. Compare a transparent baseline with the advanced method. If they disagree, investigate the mechanism before deciding which answer is more credible.

This chapter belongs to a larger decision, so translate the number carefully. State one action the evidence permits and one action it does not. A parent or educator can ask which school ideas are doing the work: algebra organises relationships, probability represents uncertainty, calculus describes local change, and computing repeats the documented rules at scale.

Contents · Previous chapter · Next chapter

Treat grouped versus individual data as a model-checking problem rather than a vocabulary exercise. The formula uses event counts and risk sets, so grouped times need correct d_j and n_j. Inspection-rounded failures can share an event time. Recompute the example from the raw values and label every unit. Interval censoring needs different estimators. Match the data structure.

The strongest check uses evidence the model did not choose for itself. Hold out later observations, an independent simulation set or a known-answer benchmark. Report errors where decisions are sensitive, especially near the failure boundary, instead of relying on one global accuracy score. A method that performs well on easy safe cases can still fail at the rare cases that motivated the analysis.

Now change one assumption and make a directional prediction before running the calculation. If the result moves the opposite way, inspect the sign convention, transformation, density ratio or constraint. This prediction-first routine develops transferable problem-solving skills because it connects symbolic work to a testable expectation.

Contents · Previous chapter · Next chapter

The mathematics of clustered systems becomes useful only when its mechanism is explicit. Greenwood’s ordinary formula assumes independent observational units. Consider the worked illustration: Multiple components from one machine may share environment. Preserve the original inputs, calculate step by step and compare with a second method whenever an analytic or high-budget reference exists. Treating correlated components as independent understates uncertainty. Use cluster-aware methods.

There are three uncertainties to keep apart. Sampling or numerical uncertainty comes from finite computation. Parameter uncertainty comes from estimating inputs. Model-form uncertainty comes from simplified physics, distributions or dependence. Reducing the first does not automatically reduce the other two. A credible report names all three and shows which one drives the decision.

For a practical investigation, create a small reproducible notebook or spreadsheet with a frozen data sheet, formula sheet, diagnostics sheet and decision note. Another student should be able to change one input and obtain the same update without guessing what the model meant.

Contents · Previous chapter · Next chapter

Use bootstrap comparison to practise sensitivity analysis. Resampling whole systems can provide an empirical check on Greenwood intervals. Compare transformed Greenwood and percentile-bootstrap limits at planned horizons. Before accepting the result, identify the quantity that would change the conclusion most. A bootstrap inherits censoring and sampling design assumptions. Resample at the correct unit.

A simple baseline is essential. It reveals whether the advanced method is improving accuracy or only adding complexity. Compare model calls, wall time, uncertainty and failure-mode coverage under one declared budget. Repeat stochastic calculations with several seeds; a favourable single run is not evidence of dependable efficiency.

Students can extend the example by adding correlation, a second failure mode or a modest measurement error. Explain which part of the formula changes and which remains fixed. This exercise shows how mathematics transfers from a clean textbook case to engineering and computing without pretending that the classroom model is a certified design tool.

Contents · Previous chapter · Next chapter

Build a small counterexample for simulation calibration. Simulate from a known survival and censoring process to check interval coverage and tail behaviour. The initial example is: Across 1000 repetitions, count how often true S(1000) lies inside. Then create a nearby case where the usual shortcut becomes unreliable. Simulation under one model is not universal proof. Vary censoring and sample size.

Counterexamples are positive learning tools. They reveal the conditions that make a method work and help students recognise when a software output deserves a pause. Keep the counterexample small enough to plot or enumerate. Check sign, support, bounds and conservation of probability before adding complexity.

When writing the conclusion, separate a documented fact from an interpretation and from an illustrative scenario. Cite the primary method paper for the factual mechanism, use the local calculation for the illustration, and label any recommendation as conditional on the stated model. This keeps career and engineering advice age-appropriate and honest.

Contents · Previous chapter · Next chapter

Connect decision interpretation to a real decision while protecting the boundary between calculation and judgement. Use uncertainty to distinguish an estimated threshold crossing from a supported one. A lower 95% limit below a requirement can trigger more evidence. First verify the arithmetic, then translate it into an allowed action and an action the evidence does not support. An interval alone does not set the requirement. State the acceptance rule externally.

The audit should include the original question, input provenance, failure convention, algorithm settings, convergence or stopping rule, and a sensitivity result. For an adaptive method, also preserve the sequence of choices. For an optimisation, preserve every constraint. For Bayesian updating, preserve the prior and observation model. These records turn a result into something another person can challenge and improve.

A final transfer question is useful: where else would the same mathematical idea appear? Density ratios appear in survey correction and machine learning, probabilistic constraints in planning, and conditional probability in medical testing. The context changes, but transparent assumptions and reproducible checks remain valuable.

Contents · Previous chapter · Next chapter

Create three event times, compute Kaplan–Meier multipliers, Greenwood terms, SE and a transformed interval. Remove one late survivor and predict the change. This is the central mathematical mechanism in practice activity. Write the quantities and assumptions before using software, then reproduce the calculation or geometric argument in a small example. Toy data are for learning. Audit every denominator. The useful habit is to keep the physical question, the probability model and the numerical method in separate columns so that a correct computation is not mistaken for complete evidence.

Did You Know? A result can be numerically stable and still answer the wrong question if the failure event, denominator or transformation changes. Test one nearby case and predict the direction of change before recalculating. Save inputs, intermediate values and units so a second reader can audit the path.

For practice, turn the example into a sensitivity check. Change one defensible input while holding the others fixed, compare the new result with the original and explain why it moved. Then construct a counterexample in which the shortcut fails. End with two sentences: what the mathematics supports, and what remains uncertain.

Contents · Previous chapter · Next chapter

Start with the operational question behind parent and educator guidance. Greenwood’s formula connects fractions, products, squares, roots and uncertainty. In the illustrative case, Ask why late events among few systems count more. Draw the relevant graph, density, boundary or timeline before calculating; the picture should use the same variables and sign convention as the equations. Begin with one event term before cumulative sums. Keep a risk-set table visible.

A student should be able to rebuild the result without hidden settings. Record the sampling rule or optimisation criterion, the exact failure definition, and the evidence used for every distribution or parameter. Compare a transparent baseline with the advanced method. If they disagree, investigate the mechanism before deciding which answer is more credible.

This chapter belongs to a larger decision, so translate the number carefully. State one action the evidence permits and one action it does not. A parent or educator can ask which school ideas are doing the work: algebra organises relationships, probability represents uncertainty, calculus describes local change, and computing repeats the documented rules at scale.

Contents · Previous chapter · Next chapter

Treat common misconceptions as a model-checking problem rather than a vocabulary exercise. Greenwood variance is not the variance of raw lifetimes, a pointwise interval is not a whole-curve band, and censoring is not an event. Narrow early intervals do not guarantee a stable tail. Recompute the example from the raw values and label every unit. Clipping invalid bounds is not a principled fix. Name scale and coverage target.

The strongest check uses evidence the model did not choose for itself. Hold out later observations, an independent simulation set or a known-answer benchmark. Report errors where decisions are sensitive, especially near the failure boundary, instead of relying on one global accuracy score. A method that performs well on easy safe cases can still fail at the rare cases that motivated the analysis.

Now change one assumption and make a directional prediction before running the calculation. If the result moves the opposite way, inspect the sign convention, transformation, density ratio or constraint. This prediction-first routine develops transferable problem-solving skills because it connects symbolic work to a testable expectation.

Continue with the Mathematics Learning Hub, the eduKate Mathematics Learning System, the approach to learning mathematics, the Singapore Mathematics curriculum overview, the cross-subject learning approach and the previous reliability article. Each destination has its own role: subject navigation, learning method, school-to-advanced progression, or continuity with the latest reliability mechanism.

Contents · Previous chapter · Next chapter

The mathematics of research and next reading becomes useful only when its mechanism is explicit. Greenwood’s 1926 report introduced the variance idea later used with Kaplan–Meier’s product-limit estimator for incomplete observations. Consider the worked illustration: Reproduce the variance sum before selecting software defaults. Preserve the original inputs, calculate step by step and compare with a second method whenever an analytic or high-budget reference exists. Historical formulas do not fix informative censoring. Read the original estimator beside modern interval guidance.

There are three uncertainties to keep apart. Sampling or numerical uncertainty comes from finite computation. Parameter uncertainty comes from estimating inputs. Model-form uncertainty comes from simplified physics, distributions or dependence. Reducing the first does not automatically reduce the other two. A credible report names all three and shows which one drives the decision.

For a practical investigation, create a small reproducible notebook or spreadsheet with a frozen data sheet, formula sheet, diagnostics sheet and decision note. Another student should be able to change one input and obtain the same update without guessing what the model meant.

Quick answers

What should I calculate first? Start with the estimand or failure event, then a simple baseline.

Does the advanced method guarantee accuracy? No. It needs input, numerical and model checks.

Which school mathematics matters? Algebra, graphs, probability, functions and careful units all contribute.

How can a student practise safely? Use fictional or open benchmark data and preserve every intermediate value.

What makes a career claim responsible? Describe how the mathematics is used without promising admission, salary or employment.

Contents · Previous chapter · Mathematics Learning Hub