Multicollinearity: Why Your Model Is Failing and How to Fix It

Multicollinearity Why Your Model Fails and How to Fix It for Dissertation | UK, US, Canada VIF above10, inflated SEs, flipping signs in MSc dissertation - need detect + fix without killing theory? You are in the right place. WHAT YOU WILL LEARN: • Detect: correlation above0.8, VIFabove10, tolerance below0.1 • Stata/EViews: estat vif, correlation matrix • Fix: drop/combine, index, ridge, theory-first choice • Common mistakes: auto-dropping key theory variable, ignoring centered interaction WHO THIS IS FOR: MSc, PhD Economics, Finance, Banking students in UK, US, Canada, Australia, EU. Hi, I'm Dr Bosede Ngozi Adeleye, Senior Lecturer in Economics (University of Lincoln, UK) and Founder of CrunchEconometrix. OFFICIAL LINKS & COMMUNITY: Website: https://cruncheconometrix.com Data Shop: https://cruncheconometrix.com/view/da... Members-Only:    • Members-only videos   Join Membership:    / @cruncheconometrix   Facebook:   / cruncheconometrix   LinkedIn:   / cruncheconometrix   YouTube:    / @cruncheconometrix   GitHub: https://github.com/CrunchEconometrix Twitter: https://x.com/crunchmetrix Teachable (600+ students since 2020, closed 30/9/2026): https://cruncheconometrix.teachable.com Subscribe: https://www.youtube.com/c/CrunchEcono... Are you writing your MSc Economics, Finance, Banking or PhD dissertation and your supervisor said multicollinearity is problem - causes and treatment - and you need to understand multicollinearity causes and treatment for your dissertation? You are in the right place. WHAT YOU WILL LEARN: ✓ Treatment of multicollinearity: Drop one of highly correlated variables - e.g., if FDI and trade correlation 0.9, drop one - theory guides which to drop - e.g., keep theoretically more important, Combine variables - create index via PCA - e.g., governance indicators correlated - combine into governance index via PCA - reduces multicollinearity, Increase sample size - more observations reduces multicollinearity - collect more data, Centering - for interaction and quadratic - center X before creating X^2 or X1*X2 - reduces correlation between X and X^2 or X1 and X1*X2, Ridge regression or principal component regression - advanced - shrinks coefficients - but for MSc dropping or combining or centering often enough - I show practical ✓ How to report multicollinearity causes and treatment results in academic format for UK/US dissertation - table with correlation matrix and VIF values - e.g., VIF for X1=2.1, X2=3.5, X3=1.8 all below 10 no serious multicollinearity - plus statement about causes - e.g., high correlation between FDI and trade due to theory both measure openness - treatment dropped trade kept FDI based on theory - what examiners want ✓ Common mistakes MSc students make: Not testing multicollinearity at all - examiners check correlation and VIF - need to report, Dropping variables without theory - need theory justification which variable to drop - not just based on VIF alone, Confusing multicollinearity with heteroskedasticity or autocorrelation - multicollinearity correlation among X's, heteroskedasticity non-constant error variance, autocorrelation correlation among errors - different, Using VIF above 5 as strict cutoff - VIF above 10 common cutoff but VIF above 5 moderate - need to discuss - if VIF 6-10 moderate but okay if theory supports WHO THIS IS FOR: MSc, MBA, PhD Economics, Finance, Banking, Management, Business Analytics students in UK (Warwick, Manchester, Leeds, Birmingham, Glasgow, Edinburgh, LSE), US, Canada, Australia, EU with multiple explanatory variables - regressors - in regression model. If this helped, LIKE, COMMENT your VIF values and country, SHARE. FAQ: Q: What causes multicollinearity? A: High correlation between explanatory variables - e.g., including same variable twice, lag and lead of same variable, dummy variable trap including all dummies and constant, interaction without centering, small sample, natural correlation - e.g., FDI and trade both measure openness correlated. Q: How to treat multicollinearity? A: Drop one of highly correlated variables - theory guides which - combine via PCA index - e.g., governance indicators into index, increase sample size, centering for interaction and quadratic, ridge regression advanced - for MSc dropping or combining or centering often enough. 0:00 Addressing Technical Challenges 1:45 Defining Multicollinearity 4:18 Spotting the Signs 5:13 Diagnostic Tools Explained 7:13 Strategies for Treatment 10:34 Practical Analysis in Stata and EViews #Multicollinearity #Causes #Treatment #CrunchQueenSpace #CQS #VIF #MulticollinearityCauses #MulticollinearityTreatment #Stata #EViews #Econometrics #StataTutorial #MScDissertation #UKUniversities #VIFTest #CorrelationMatrix #PhDResearch #DissertationHelp #CQSData #Dofile #CrunchEconometrix