Where does fit collide?

In statistics, goodness of fit is how closely a fitted model or distribution reproduces observed data, expressed as the discrepancy between what the model predicts and what was seen. Pearson's 1900 chi-square test compared observed and expected frequencies to judge whether a distribution could have generated a sample, and modern structural equation modelling reports indices such as RMSEA and CFI, whose thresholds Hu and Bentler examined in 1999. The referent is a model-data pair, and a small discrepancy, or an index past its cutoff, signals a fit. Nothing here is a person, a firm, or a mask: fit is a numerical property of how well equations track evidence.[1][2]

Unit

The discrepancy between a model and observed data

Recovery means

Not applicable

Measured by

Chi-square goodness-of-fit statistics, and SEM indices such as RMSEA and CFI against cutoffs

In organizational behavior, person-environment fit is the compatibility between an individual and their work setting, whether the two supply each other's needs or share the same attributes. Kristof's 1996 review sorted the idea into person-job and person-organisation fit, and into supplementary fit, where person and setting are alike, and complementary fit, where each meets the other's demands; Kristof-Brown, Zimmerman, and Johnson's 2005 meta-analysis pooled the evidence linking fit to satisfaction and commitment. The referent is a person set against a job, organisation, or group, and the two sides are values, needs, and abilities, not model equations. This is a congruence between people and workplaces, not a statistic and not a seal.[3][4]

Unit

A person paired with a job, organisation, or group

Recovery means

Not applicable

Measured by

Direct perceived-fit items, indirect difference scores, and polynomial regression

In strategy and contingency theory, fit is the alignment between an organisation's structure or strategy and its context that is held to drive performance: the structure suited to a given environment, technology, or strategy outperforms a mismatched one. Drazin and Van de Ven set out in 1985 how contingency research models this alignment as selection, interaction, or systems fit, and Venkatraman's 1989 review distinguished six ways of specifying fit, from moderation and mediation to matching, profile deviation, covariation, and gestalts, warning that the verbal claim and the statistical test often diverge. The referent is a whole organisation set against its situation, not a person set against a job. It shares the word with person-environment fit, but the fitted pair is a firm and its environment.[5][6]

Unit

An organisation's structure or strategy set against its context

Recovery means

Realignment after a shift in strategy or environment

Measured by

Fit as moderation, matching, profile deviation, or covariation between structure and context

In occupational medicine, fit is how well a tight-fitting respirator seals against the wearer's face, so that air passes through the filter rather than leaking around the edges. It is quantified by fit testing as a fit factor, the ratio of a contaminant's concentration outside the mask to its concentration inside, measured on a specific person wearing a specific facepiece. Coffey and colleagues in 2004 tested eighteen N95 filtering-facepiece respirators across many subjects and showed that measured fit depends on the pairing of model and face. The referent is a physical seal on one body, counted as a leakage ratio, and it has nothing to do with a model's fit to data or a firm's fit to its market.[7]

Unit

The seal between a respirator and one wearer's face, as a fit factor

Recovery means

Not applicable

Measured by

Fit testing: the fit factor, the ratio of ambient to in-mask contaminant concentration

In design theory, fit refers to the fit between a designed form and its context, judged by the absence of misfits between them. This meaning comes from Alexander (1964) in Notes on the synthesis of form.[8]

Unit

Recovery means

Measured by