New functionality: Diagnostics for linear regression models

ByTrond Pedersen

It is now possible to perform tests and diagnostics of linear regression models in microdata.no through the use of options. Examples are tests for heteroskedasticity, multicollinearity and omitted variables.

  • ov: Ramsey’s RESET test for omitted variables. Displays a total F-value with associated P-value
  • vif: Variance inflation factor test for multicollinearity. Displays variance inflation factor (VIF) values for the independent variables in a table
  • het_bp: Breusch-Pagan test for heteroskedasticity. Displays a total chi-square value with associated P-value
  • het_iid: Studentized Breusch-Pagan test for heteroskedasticity. Displays an overall chi-square value with associated P-value (a newer version of the BP test that is more robust since it does not assume that the residuals are normally distributed)
  • het_fstat: F-statistics from the Breusch-Pagan test for heteroskedasticity. Displays a total F-value with associated P-value. The number of degrees of freedom is not based on the number of variables in the regression model, but stems from an additional OLS model that compares residuals and predicted values, i.e. only one degree of freedom

Example of regression analysis including tests for omitted variables, multicollinearity and heteroscedasticity:

regress income2004 sex income2003, ov vif het_bp

Microdata.no offers several more options for model diagnostics:

  • The command correlate provides pairwise measures of correlation between individual variables in a table matrix
  • The command regress-predict provides the opportunity to study e.g. residual and prediction values. This can be combined with graphical tools such as histograms to show residual distributions and to check for e.g. normal distribution. The hexbin command can also display an anonymized 2-way plot. Click here for more info

We work continuously to improve microdata.no so that most analysis needs can be met. Thus, more and more analysis options will be introduced. We are based on input from researchers, and have compiled a list of new tools that we work through. Do you have specific suggestions for new functionality? Feel free to contact us by email: support@microdata.no