Linear model: Difference between revisions

Revision as of 11:13, 5 January 2014

Latest revision as of 23:29, 17 November 2024

(38 intermediate revisions by 29 users not shown)

Line 1:

{{Short description|Type of statistical model}}

{{distinguishDistinguish|linear model of innovation}}

In [[statistics]], the term '''linear model''' isrefers usedto inany differentmodel wayswhich accordingassumes to[[linearity]] in the contextsystem. The most common occurrence is in connection with regression models and the term is often taken as synonymous with [[linear regression]] model. However, the term is also used in [[time series analysis]] with a different meaning. In each case, the designation "linear" is used to identify a subclass of models for which substantial reduction in the complexity of the related [[statistical theory]] is possible.

{{merge|Linear predictor function|discuss=Talk:Linear predictor function#Proposed merge with Linear model|date=January 2014}}

In [[statistics]], the term '''linear model''' is used in different ways according to the context. The most common occurrence is in connection with regression models and the term is often taken as synonymous with [[linear regression]] model. However, the term is also used in [[time series analysis]] with a different meaning. In each case, the designation "linear" is used to identify a subclass of models for which substantial reduction in the complexity of the related [[statistical theory]] is possible.

==Linear regression models==

{{main|Linear regression}}

For the regression case, the [[statistical model]] is as follows. Given a (random) sample <math> (Y_i, X_{i1}, \ldots, X_{ip}), \, i = 1, \ldots, n </math> the relation between the observations ''Y<submath>iY_i</submath>'' and the [[independent variables]] ''X<submath>X_{ij}</submath>'' is formulated as

:<math>Y_i = \beta_0 + \beta_1 \phi_1(X_{i1}) + \cdots + \beta_p \phi_p(X_{ip}) + \varepsilon_i \qquad i = 1, \ldots, n </math>

where <math> \phi_1, \ldots, \phi_p </math> may be [[Nonlinear system|nonlinear]] functions. In the above, the quantities ''ε<submath>i\varepsilon_i</submath>'' are [[random variablesvariable]]s representing errors in the relationship. The "linear" part of the designation relates to the appearance of the [[regression coefficient]]s, ''β<submath>j\beta_j</submath>'' in a linear way in the above relationship. Alternatively, one may say that the predicted values corresponding to the above model, namely

:<math>\hat{Y}_i = \beta_0 + \beta_1 \phi_1(X_{i1}) + \cdots + \beta_p \phi_p(X_{ip}) \qquad (i = 1, \ldots, n), </math>

are linear functions of the ''β<submath>j\beta_j</submath>''.

Given that estimation is undertaken on the basis of a [[least squares]] analysis, estimates of the unknown parameters ''β<submath>j\beta_j</submath>'' are determined by minimising a sum of squares function

:<math>S = \sum_{i = 1}^n \varepsilon_i^2 = \sum_{i = 1}^n \left(Y_i - \beta_0 - \beta_1 \phi_1(X_{i1}) - \cdots - \beta_p \phi_p(X_{ip})\right)^2 .</math>

From this, it can readily be seen that the "linear" aspect of the model means the following:

:*the function to be minimised is a quadratic function of the ''β<submath>j\beta_j</submath>'' for which minimisation is a relatively simple problem;

:*the derivatives of the function are linear functions of the ''β<submath>j\beta_j</submath>'' making it easy to find the minimising values;

:*the minimising values ''β<submath>j\beta_j</submath>'' are linear functions of the observations ''Y<submath>iY_i</submath>'';

:*the minimising values ''β<submath>j\beta_j</submath>'' are linear functions of the random errors ''ε<submath>i\varepsilon_i</submath>'' which makes it relatively easy to determine the statistical properties of the estimated values of ''β<submath>j\beta_j</submath>''.

==Time series models==

An example of a linear time series model is an [[autoregressive moving average model]]. Here the model for values {''X<submath>tX_t</submath>''} in a time series can be written in the form

:<math> X_t = c + \varepsilon_t + \sum_{i=1}^p \varphi_iphi_i X_{t-i} + \sum_{i=1}^q \theta_i \varepsilon_{t-i}.\,</math>

where again the quantities ''ε<submath>t\varepsilon_i</submath>'' are random variables representing [[Innovation (signal processing)|innovations]] which are new random effects that appear at a certain time but also affect values of ''<math>X''</math> at later times. In this instance the use of the term "linear model" refers to the structure of the above relationship in representing ''X<submath>tX_t</submath>'' as a linear function of past values of the same time series and of current and past values of the innovations.<ref>Priestley, M.B. (1988) ''Non-linear and Non-stationary time series analysis'', Academic Press. {{ISBN |0-12-564911-8}}</ref> This particular aspect of the structure means that it is relativerelatively simple to derive relations for the mean and [[covariance]] properties of the time series. Note that here the "linear" part of the term "linear model" is not referring to the coefficients ''φ<submath>i\phi_i</submath>'' and ''θ<submath>i\theta_i</submath>'', as it would be in the case of a regression model, which looks structurally similar.

==Other uses in statistics==

There are some other instances where "nonlinear model" is used to contrast with a linearly structured model, although the term "linear model" is not usually applied. One example of this is [[nonlinear dimensionality reduction]].

==See also==

* [[General linear model]]

* [[Generalized linear model]]

* [[Linear predictor function]]

* [[Linear system]]

* [[Linear regression]]

* [[Statistical model]]

==References==

{{Reflist}}

<references/>

{{Statistics}}

{{Authority control}}

[[Category:StatisticalCurve modelsfitting]]

[[Category:Time seriesRegression models]]

[[Category:Regression analysis]]

[[Category:Statistical terminology]]

[[ar:نموذج الانحدار الخطي]]

[[de:Lineares Modell]]

[[fr:Modèle linéaire]]