Where does replacement collide?

In computing, a replacement policy is the rule a cache or virtual-memory system uses to decide which stored item to throw out when it must make room for a new one, so here "replacement" names the swapping-out of a page or cache line rather than any renewal of a population. Belady's 1966 A study of replacement algorithms for a virtual-storage computer compared strategies such as least-recently-used against the theoretically optimal policy that evicts the item needed furthest in the future. The referent is an eviction decision inside a storage hierarchy, judged by hit and miss rates, and it is unrelated to a fertility threshold or to returning a sampled unit to an urn.[1]

Unit

The choice of which cached item to evict, scored by the resulting hit/miss rate

Recovery means

Not applicable

Measured by

Evaluated by cache hit ratio and page-fault rate under a given reference stream and policy

In probability and statistics, sampling with replacement means each unit drawn is returned to the population before the next draw, so the same unit can be selected more than once and successive draws stay independent and identically distributed; sampling without replacement does the opposite. Feller's An Introduction to Probability Theory and Its Applications develops the urn models that make the distinction precise, and the scheme underlies resampling methods such as the bootstrap. The referent is a drawing rule that fixes the probability structure of a sample, not an eviction decision or a demographic benchmark.[2]

Unit

Whether a drawn unit is returned to the pool before the next draw

Recovery means

Not applicable

Measured by

Specified in the sampling design; it determines whether draws are independent and whether units may repeat

In demography, replacement-level fertility is the number of births per woman at which a generation exactly replaces itself over the long run, conventionally about 2.1 children in low-mortality populations, just above two because some girls do not survive to reproduce. Espenshade, Guzman, and Westoff's 2003 The surprising global variation in replacement fertility showed that the exact threshold rises with mortality, from near 2.1 in rich countries to well above 3 where child survival is low. The referent is a fertility benchmark for population stability, compared against the total fertility rate, and it has nothing to do with evicting a cache page or returning a unit to a sampling urn.[3]

Unit

Births per woman required for a generation to replace itself (≈2.1 at low mortality)

Recovery means

Not applicable

Measured by

Derived from the total fertility rate and female survival to the mean age of childbearing, then compared with the observed TFR