Scribd Logo
Upload

Welcome to Scribd!

  • 19 views

    Esearch Ethods in Inance: Multicollineraity

    Uploaded by

    Usman Baloch
    The document discusses multicollinearity in regression analysis. It defines multicollinearity as existing when explanatory variables are highly correlated with each other. It lists the effects of multicollinearity as unstable beta coefficients, incorrect sign on coefficients, unstable t-statistics and F-statistics. Two methods are presented for detecting multicollinearity: examining the correlation matrix between variables and calculating auxiliary R-squared values from regressions of each variable against the others. Tolerance limits using the variance inflation factor are suggested to determine if highly correlated variables can still be used together in a regression.

    Copyright:

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

    Esearch Ethods in Inance: Multicollineraity

    Uploaded by

    Usman Baloch
    19 views4 pages
    The document discusses multicollinearity in regression analysis. It defines multicollinearity as existing when explanatory variables are highly correlated with each other. It lists the effects of multicollinearity as unstable beta coefficients, incorrect sign on coefficients, unstable t-statistics and F-statistics. Two methods are presented for detecting multicollinearity: examining the correlation matrix between variables and calculating auxiliary R-squared values from regressions of each variable against the others. Tolerance limits using the variance inflation factor are suggested to determine if highly correlated variables can still be used together in a regression.

    Original Title

    Lec 7, Sec 1, 2 & 3

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses multicollinearity in regression analysis. It defines multicollinearity as existing when explanatory variables are highly correlated with each other. It lists the effects of multicollinearity as unstable beta coefficients, incorrect sign on coefficients, unstable t-statistics and F-statistics. Two methods are presented for detecting multicollinearity: examining the correlation matrix between variables and calculating auxiliary R-squared values from regressions of each variable against the others. Tolerance limits using the variance inflation factor are suggested to determine if highly correlated variables can still be used together in a regression.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    19 views4 pages

    Esearch Ethods in Inance: Multicollineraity

    Uploaded by

    Usman Baloch
    The document discusses multicollinearity in regression analysis. It defines multicollinearity as existing when explanatory variables are highly correlated with each other. It lists the effects of multicollinearity as unstable beta coefficients, incorrect sign on coefficients, unstable t-statistics and F-statistics. Two methods are presented for detecting multicollinearity: examining the correlation matrix between variables and calculating auxiliary R-squared values from regressions of each variable against the others. Tolerance limits using the variance inflation factor are suggested to determine if highly correlated variables can still be used together in a regression.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 4

    Page 1 of 4

    Research Methods in Finance

    Abdul Qadeer Khan (PhD Scholar & Research Associate)


    nd
    D-Block 2 Floor, Room # 218
    Mohammad Ali Jinnah University, Islamabad
    Office: 051-4486701 Ext: 212
    Cell: 0333-6487274
    [email protected]; [email protected]

    Section 1, 2 & 3
    Lecture # 7
    July 17-20, 2012
    Multicollineraity:

    If there is significant relationship between explanatory variable(s), then the issue of multicollineraity can
    exit. Y = α + βX1 + βX2 , in this equation if the variables X1 and X2 explain the same phenomena, the issue
    of multicollineraity in variables exists.
    Page 2 of 4

    Effects of Multicollineraity

    Effects of Multicollineraity are:

    i. β : If sample size increase or decrease the beta sign will change. Beta is unstable, sign
    incorrect and sign instable. For example if we take the sample of five years the beta sign will
    be different to the sample size of ten years.
    ii. Standard Error
    iii. T-Statistics
    iv. F-Statistics

    Detection of Multicollineraity:

    Multicollineraity can be detected by two ways:

    i. Correlation Efficient (Correlation Matrix)


    ii. Auxiliary R2

    a. Correlation Efficient (Correlation Matrix)

    Follow these steps:

    i. Generate variables (shortcut method for putting data into eviews: note observations of data
    and open eviews, put observations with unstructured frequency and click OK. Now copy from
    excel all variables with series name and paste in small window of eviews, then follow the
    steps and finish)
    ii. Go to quick menu-----Group statistics and click Correlation
    iii. In new window, put variable names (without intercept) as: y x1 x2
    iv. Some results will display:
    Page 3 of 4

    Interpretation:

    In above results, we will check (even read) the correlation matrix as: X1 relation with X2 [here first
    variable X1 shows row and second X2 shows as column] so, the correlation of X1 with X2 is -.149838. We
    take subjective decision that if answer is more than .5 then the chances of multicollineraity will greater.

    Auxiliary R2

    This is the second way to detect multicollineraity. First, regress basic equation for initial idea about R2 as
    y c x1 x2 and observe R2 and then regress:

    y c x1 --------------------Observe and note R2 `----Eq.1

    y c x2 -------------------- Observe and note R2 ----Eq.2

    As closest the answer of R2 with each other, chances for the existence of multicollineraity will increase. If
    you observe that the answers of R2 by regressing above equations separately are close to each other, it
    known that chances of multicollineraity are greater so for confirmation about multicollineraity we
    regress equation as: X1 c x2 and if the answers of R2 is near to 0.8 or 0.9 then multicollineraity exists.
    Which Variable as Dependant Variable? Use dependent variable only on the basis of economic rationale,
    Page 4 of 4

    reasoning and economic locality. Suppose there are two variables population and GDP, here GDP will be
    used as dependent variable and population as explanatory variable.

    There are two ways to deal with multicollineraity:

    i. First way: we cannot use alongwith


    ii. If we want to use both, then we test whether these variables are inside the tolerable limit or
    not. If VIF < 5; then it indicates that these two variables are in tolerable limit and can be
    used alongwith otherwise we will have to sacrifice one.

    Formula of VIF (Variance Inflationary Factor):


    VIF = 1/ 1- Adj. R2 ----- (Here pick Adj. R2 by regressing y c x1 x2)

    Decision Criteria:

    If VIF < 5 then we consider it on the basis of different literature that this is in-tolerable limit
    and can be used both variable.

    If VIF > 5 then we consider it on the basis of different literature that this is out-of-tolerable
    limit and one variable must be sacrificed.

    You might also like