Where does link collide?

A link is a connection between two actors in a social network, a tie that records a relationship such as friendship, advice, or communication. Jacob Moreno's 1934 Who Shall Survive? founded sociometry and drew the first sociograms, mapping people as points and their relationships as the lines between them, and Wasserman and Faust's 1994 Social Network Analysis: Methods and Applications gathered the field's methods and treated these ties, also called links, as the basic relational unit joining nodes. Here a link is a social or relational connection between actors, present or absent and sometimes weighted or directed, not a measured length on the ground, a reference between documents, or a function in a regression model.[1][2]

Unit

One tie between two actors, recorded as present or absent and sometimes given a weight or a direction

Recovery means

Not applicable

Measured by

Observed from surveys, rosters, or interaction records and entered as edges in an adjacency matrix

A hyperlink, or link, is a reference embedded in one document that a reader can follow to reach another document or another place within the same one. Ted Nelson's 1965 A File Structure for the Complex, the Changing and the Indeterminate coined the word hypertext and described writing joined by such links, and Berners-Lee and colleagues' 1994 The World-Wide Web described the global hypertext system in which a link is an address that a browser resolves and retrieves on request. Here a link is a navigable pointer between pieces of digital content, defined by its source, its target address, and the act of following it, not a length on the ground, a tie between people, or a transformation in a statistical model.[3][4]

Unit

One directed reference from a source anchor to a target address that a reader can follow

Recovery means

Not applicable

Measured by

Counted and traced through a document graph by source, target, and whether the target address resolves

In a generalized linear model the link function is the transformation that connects the linear predictor, a weighted sum of the explanatory variables, to the mean of the response. Nelder and Wedderburn's 1972 Generalized Linear Models introduced this framework and named the link function, showing how one method fits models as different as ordinary regression, logistic regression, and Poisson regression by choosing a different link, such as the identity, the logit, or the log. Here a link is a mathematical function chosen by the analyst to relate a predictor to an expected value, not a length on the ground, a tie between people, or a reference between documents.[5]

Unit

One monotonic function g relating the mean response to the linear predictor, so that g(mean) equals the predictor

Recovery means

Not applicable

Measured by

Chosen in advance to suit the response distribution and checked by model fit and residual behaviour