Find your next learning step
A polymer domain may leave a regime between two microscope frames. Replacing that interval with its midpoint creates a precise-looking time the experiment never observed and can distort which mechanism appears fastest.
Open the chapter contents · Polymer Science root · Gold Standard of Education · How X Works
Chapter contents
Choose the route closest to the question, or read from Chapter 1 for the complete argument. The chapters remain open on the page; these grouped routes are an index.
Route 1 · Frame the problem
Route 2 · Work the ledgers
Route 3 · Transfer the method
Route 4 · Check limits and evidence
Route 5 · Ask and continue
CHAPTER 1 OF 12 · FRAME THE PROBLEM
1. The event happened inside an interval, not at a guessed instant
Suppose a polymer specimen is inspected at scheduled times. It is still in the original regime at 20 minutes and has exited by 30 minutes. The data say 20 Interval censoring differs from ordinary right censoring. A right-censored specimen is known only to survive beyond its last observation; an interval-censored exit is bracketed by two observations. Exact-time survival tools applied to guessed midpoints can bias timing, uncertainty and protocol comparisons.
CHAPTER 2 OF 12 · FRAME THE PROBLEM
2. Inspection schedules shape the available timing resolution
A fixed ten-minute schedule produces wider intervals than a two-minute schedule. An adaptive schedule may inspect unstable specimens more often, making interval width depend on emerging behaviour. The observation process then carries information and must be documented or modelled.
Schedule differences can mimic mechanism differences. If protocol A is inspected every two minutes and protocol B every ten, A will appear to have more precisely timed and sometimes earlier exits even when the latent event distributions are identical.
CHAPTER 3 OF 12 · WORK THE LEDGERS
3. A six-specimen interval ledger
Consider six tracked domains under one protocol. The first four exit between visits; one remains in regime at the final visit; one has already exited before the first usable observation.
| Domain | Last known in regime L | First known out R | Observation type |
|---|---|---|---|
| D1 | 10 | 20 | Interval-censored |
| D2 | 20 | 30 | Interval-censored |
| D3 | 20 | 40 | Interval-censored after missed frame |
| D4 | 40 | 50 | Interval-censored |
| D5 | 60 | ∞ | Right-censored |
| D6 | 0 | 10 | Left-censored within start window |
The likelihood contribution for an interval exit is the probability that T lies in its observed interval, such as F(R)−F(L). A right-censored contribution is survival beyond L. The missed frame for D3 honestly widens its timing information.
CHAPTER 4 OF 12 · WORK THE LEDGERS
4. How midpoint substitution changes a mechanism comparison
Suppose exits are classified as scission or disentanglement. Scission is inspected every five minutes because it produces a strong signal; disentanglement is checked every twenty minutes because it requires a slower assay. Midpoints may make scission appear earlier even when its intervals overlap the disentanglement intervals.
| Mechanism label | Observed intervals | Midpoint summary | What remains unknown |
|---|---|---|---|
| Scission | (10,15], (15,20], (20,25] | 15, 17.5, 22.5 | Ordering within and across intervals |
| Disentanglement | (0,20], (20,40], (20,40] | 10, 30, 30 | Broad early and later timing |
| Comparison | Different visit grids | Scission median looks smaller | How much is grid-induced |
An interval likelihood uses each bracket and can incorporate the visit process. It may still find a difference, but the conclusion is tied to what was observed rather than an invented clock time.
CHAPTER 5 OF 12 · WORK THE LEDGERS
5. Competing mechanisms need joint timing and cause bookkeeping
When the first exit can occur by several mechanisms, each specimen has a latent event time and cause. Cumulative incidence for one cause depends on the hazards of all causes because a different first exit removes the specimen from experiencing that cause first.
If the cause is uncertain as well as the time, do not solve the problems separately by filling in a midpoint and then assigning a hard label. A likelihood or multiple-imputation analysis should carry both interval and cause uncertainty, with validation evidence for the classification process.
CHAPTER 6 OF 12 · TRANSFER THE METHOD
6. Nonparametric and regression routes
Turnbull-type nonparametric estimation respects arbitrary censoring intervals for a single event distribution. For competing mechanisms and protocol effects, semiparametric models can target cause-specific quantities or cumulative incidence, but their assumptions and estimands differ.
A proportional subdistribution-hazard coefficient is not a direct physical rate constant. A cause-specific hazard answers an instantaneous transition question among specimens still free of any exit; cumulative incidence answers the probability of a particular first exit by time t. Report the quantity actually fitted.
CHAPTER 7 OF 12 · TRANSFER THE METHOD
7. Changed-context transfer: accelerated ageing
Transfer the workflow to polymer coating failure observed during accelerated ageing. Cracking, delamination and chemical embrittlement are competing first exits, while specimens are inspected at chamber openings. Each failure time is bracketed by visits and the schedule may differ by temperature.
The transfer check asks whether visit timing was prespecified, whether destructive measurements create additional censoring and whether temperature changes detectability as well as physics. A model that ignores those changes can turn measurement frequency into an apparent acceleration factor.
CHAPTER 8 OF 12 · TRANSFER THE METHOD
8. Connecting timing to existing mechanism guides
The competing-exit guide separates cause-specific hazards from cumulative incidence. The uncertain-cause guide carries label ambiguity through the mechanism ledger. This article adds uncertainty about when the exit occurred.
Use the Polymer Science root to reconnect event-time inference to chain architecture, thermal history, microscopy and spectroscopy. Timing models do not establish a molecular mechanism without appropriate physical evidence.
CHAPTER 9 OF 12 · CHECK LIMITS AND EVIDENCE
9. Limits and misconceptions
Interval-censoring methods do not recover exact hidden event times. They estimate a distribution or regression relation consistent with the brackets and assumptions. Narrow intervals are not automatically unbiased if the schedule responds to early warning signals.
More frequent inspection can perturb the specimen, and missed visits may be related to instability. Treating all censoring as independent can fail. A smooth fitted curve does not mean the underlying transition is gradual; the event definition and observation rule still matter.
CHAPTER 10 OF 12 · CHECK LIMITS AND EVIDENCE
10. Practical evidence checklist
Archive every planned and actual visit time, the last confirmed in-regime observation, the first confirmed out-of-regime observation, missed frames, end-of-follow-up rule, cause evidence and reviewer uncertainty. Keep the original brackets rather than only imputed times.
Compare protocols on a common estimand, show interval-width distributions and repeat the analysis with plausible observation-process models. Report cumulative incidence or survival with uncertainty, sensitivity to missed visits, and results from any midpoint analysis only as a labelled contrast, not the primary evidence.
CHAPTER 11 OF 12 · ASK AND CONTINUE
11. Frequently asked questions
Can a very narrow interval be treated as exact?
It may be a useful approximation, but preserve the bracket and check that the approximation is negligible relative to the scientific contrast.
Is the midpoint unbiased?
Not in general. Its behaviour depends on the event distribution and inspection process; using it also understates timing uncertainty.
What if mechanism is learned later by a stronger assay?
Use the later evidence as validation or probabilistic attribution, recording whether the assay selection depended on the observed trajectory.
CHAPTER 12 OF 12 · ASK AND CONTINUE
12. Sources and next reading
Turnbull develops the nonparametric distribution estimator for grouped, censored and truncated data. Li develops the Fine–Gray model under interval-censored competing risks. Mao, Lin and Zeng present semiparametric regression for interval-censored competing risks. The polymer ledgers above are original teaching examples.
Continue through the Polymer Science root, competing exit risks and probabilistic mechanism attribution. The Gold Standard of Education and How X Works provide the wider route.
Top · Contents · Previous · How X Works