Start with OLS and check the Gauss-Markov assumptions, then test for common problems and fix them: heteroskedasticity, autocorrelation, multicollinearity and non-stationarity, with panel data needing a choice between fixed and random effects. Report diagnostics alongside results and interpret coefficients in economic terms.
| Problem | Test | Typical fix |
|---|
| Heteroskedasticity | Breusch-Pagan or White test | Robust standard errors |
| Autocorrelation | Durbin-Watson or Breusch-Godfrey | Newey-West standard errors |
| Multicollinearity | Variance inflation factor (VIF) | Drop or combine variables |
| Non-stationarity | Augmented Dickey-Fuller test | Differencing; cointegration (Engle-Granger, Johansen) |
| Panel data model choice | Hausman test | Fixed or random effects |
Tutors work in Stata, EViews, R, Python (statsmodels) or Excel, using data from the ONS, Bank of England Database, OBR, HM Treasury, World Bank WDI, IMF WEO, OECD, Eurostat and FRED.