Scribd Logo
Upload

Welcome to Scribd!

  • 20 views

    Multicol

    Uploaded by

    Thane Snyman
    Multicollinearity (MC) occurs when there is an exact or near-exact linear relationship between explanatory variables in a regression model. MC inflates the variances of the regression coefficients, weakening the analysis and making inferences difficult. When explanatory variables are perfectly linearly related, the determinant of X'X approaches zero, undermining the calculation of the coefficient estimates and covariances. MC can arise from data collection methods, model constraints or specification, or having too many explanatory variables. It is detected using variance inflation factors, tolerance values, correlation between variables, and other diagnostics. Potential remedies include removing variables, transforming variables, or using alternative regression techniques like ridge regression.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Multicol

    Uploaded by

    Thane Snyman
    20 views3 pages
    Multicollinearity (MC) occurs when there is an exact or near-exact linear relationship between explanatory variables in a regression model. MC inflates the variances of the regression coefficients, weakening the analysis and making inferences difficult. When explanatory variables are perfectly linearly related, the determinant of X'X approaches zero, undermining the calculation of the coefficient estimates and covariances. MC can arise from data collection methods, model constraints or specification, or having too many explanatory variables. It is detected using variance inflation factors, tolerance values, correlation between variables, and other diagnostics. Potential remedies include removing variables, transforming variables, or using alternative regression techniques like ridge regression.

    Original Title

    Multicol (1)

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Multicollinearity (MC) occurs when there is an exact or near-exact linear relationship between explanatory variables in a regression model. MC inflates the variances of the regression coefficients, weakening the analysis and making inferences difficult. When explanatory variables are perfectly linearly related, the determinant of X'X approaches zero, undermining the calculation of the coefficient estimates and covariances. MC can arise from data collection methods, model constraints or specification, or having too many explanatory variables. It is detected using variance inflation factors, tolerance values, correlation between variables, and other diagnostics. Potential remedies include removing variables, transforming variables, or using alternative regression techniques like ridge regression.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    20 views3 pages

    Multicol

    Uploaded by

    Thane Snyman
    Multicollinearity (MC) occurs when there is an exact or near-exact linear relationship between explanatory variables in a regression model. MC inflates the variances of the regression coefficients, weakening the analysis and making inferences difficult. When explanatory variables are perfectly linearly related, the determinant of X'X approaches zero, undermining the calculation of the coefficient estimates and covariances. MC can arise from data collection methods, model constraints or specification, or having too many explanatory variables. It is detected using variance inflation factors, tolerance values, correlation between variables, and other diagnostics. Potential remedies include removing variables, transforming variables, or using alternative regression techniques like ridge regression.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 3

    STK 310 (Multicollinearity (MC))

    Some ideas:
    MC is a problem of exact or near exact linear relationship between the explanatory of feature
    variables. Not a problem in the structure of the error terms.
    Is this really a problem?

    Consider the model:

    Yi = β̂1 + β̂2 X2 + β̂3 X3 + ûi (1)


    Ŷi = β̂1 + β̂2 X2 + β̂3 X3 (2)

    or in deviation(?) form:

    yi = β̂2 x2 + β̂3 x3 + ûi (3)


    ŷi = β̂2 x2 + β̂3 x3 (4)

    It is known that:

    σ2 σ2 1
    var(β̂2 ) = P = × (5)
    x22i (1 − r23
    2 ) 2 )
    x22i (1 − r23
    P

    σ2 σ2 1
    var(β̂3 ) = P = × (6)
    x23i (1 − r232 ) x23i (1 − r23
    2 )
    P

    −r23 σ 2
    cov(β̂2 , β̂3 ) = qP (7)
    2 ) x22i
    P 2
    (1 − r23 x3i

    where r23 is the correlation coefficient between X2 and X3 .


    What happens if X2 and X3 are independent?
    What happens if X2 and X3 are perfectly linearly related?

    Matrix results (considering 1, 2 above)

    ŷ = Xβ + û (8)
    −1 0
    β̂ = X 0 X Xy (9)
    −1
    cov(β̂) = σ 2 X 0 X (10)

    In order to determine β̂ or cov(β̂) we need the determinant of X 0 X.


    What happens with |X 0 X| when X2 and X3 are perfectly linearly related?

    1
    STK 310 (Multicollinearity (MC))

    Sources of MC???? See textbook/class discussion for more detail

    1. Data collection method

    2. Constraints on the model

    3. Model specification

    4. Overdetermined model

    Consequences of MC???? See textbook/class discussion for more detail

    1. Although the OLS estimators are BLUE their variances are inflated
    Intuition ??? Impact of inflated variances?

    σ2
    var(βˆj ) = P 2 V IFj (11)
    xj
    1
    V IFj = (12)
    1 − Rj2

    where
    β̂j = the estimated partial regression coefficient of variable Xj
    Rj2 = R2 of the regression of Xj on the remaining (k − 2) explanatory variables
    P 2
    (Xj − X j )2
    P
    xj =
    1
    Sometimes we use tolerance: T OLj = V IFj .

    2. Confidence intervals tend to be much wider, eg. (β̂j ± t/alpha/2 se(β̂j ))

    3. t-ratios lower

    4. Low t-values but high R2

    5. OLS estimators and standard errors sensitive to small changes in the data.

    2
    STK 310 (Multicollinearity (MC))

    Detection of MC

    1. High R2 but few significant t ratios.

    2. High pairwise correlation between explanatory variables

    3. Examination of partial correlation

    4. Auxiliary regressions

    5. Eigenvalues and condition index

    6. Variance inflation factors (and tolerance)

    7. Plots

    Remedial measures

    1. Use a-priori information

    2. Extending/expanding data

    3. Removing variables

    4. Transformation of variables

    5. Ridge regression/ factor analysis regression etc

    You might also like

    pFad - Phonifier reborn

    Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

    Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


    Alternative Proxies:

    Alternative Proxy

    pFad Proxy

    pFad v3 Proxy

    pFad v4 Proxy