Pitfalls of regression modelling with complex survey data Florian Ertz and Ralf M¨ unnich Economic and Social Statistics Department Trier University New Techniques and Technologies for Statistics (NTTS) Brussels Session 9A – Lightning Bolt Presentations 11 March 2015
The HFCS The Problem The Solution The HFCS ◮ Wealth surveys obviously more important than ever (demographic change and global financial crisis) ◮ European Central Bank (ECB) initiated the Eurosystem Household Finance and Consumption Survey (HFCS) ◮ National central banks collect (ex ante harmonised) micro data on private households’ assets, liabilities and consumption ◮ First wave (2008–2011) covered 15 member states, 62,521 households, and 154,247 individuals ◮ Integrates oversampling of wealthy households Brussels, 11/03/2015 | Ertz and M¨ unnich | 1 (3) Pitfalls of regression modelling with complex survey data
The HFCS The Problem The Solution The Problem W D 100 1st D 100 min D 100 max D 1000 (Intercept) age you age old chil edu low edu high emp edu emp oth emp ret emp self emp un fem inc core quin ctry2 inc core quin ctry3 inc core quin ctry4 inc core quin ctry5 mar div mar mar mar wid mult earn w net quin ctry2 w net quin ctry3 w net quin ctry4 w net quin ctry5 Data source: HFCS (2013). ***/**/*: Significance at 1/5/10% level. Brussels, 11/03/2015 | Ertz and M¨ unnich | 2 (3) Pitfalls of regression modelling with complex survey data
The HFCS The Problem The Solution The Problem W D 100 1st D 100 min D 100 max D 1000 (Intercept) *** age you ** age old * chil edu low edu high ** emp edu *** emp oth emp ret ** emp self emp un fem inc core quin ctry2 ** inc core quin ctry3 * inc core quin ctry4 *** inc core quin ctry5 ** mar div mar mar mar wid *** mult earn ** w net quin ctry2 *** w net quin ctry3 *** w net quin ctry4 *** w net quin ctry5 *** Data source: HFCS (2013). ***/**/*: Significance at 1/5/10% level. Brussels, 11/03/2015 | Ertz and M¨ unnich | 2 (3) Pitfalls of regression modelling with complex survey data
The HFCS The Problem The Solution The Problem W D 100 1st D 100 min D 100 max D 1000 (Intercept) *** *** age you ** *** age old * *** chil edu low edu high ** *** emp edu *** *** emp oth ** emp ret ** *** emp self emp un fem inc core quin ctry2 ** *** inc core quin ctry3 * *** inc core quin ctry4 *** *** inc core quin ctry5 ** *** mar div mar mar mar wid *** *** mult earn ** ** w net quin ctry2 *** *** w net quin ctry3 *** *** w net quin ctry4 *** *** w net quin ctry5 *** *** Data source: HFCS (2013). ***/**/*: Significance at 1/5/10% level. Brussels, 11/03/2015 | Ertz and M¨ unnich | 2 (3) Pitfalls of regression modelling with complex survey data
The HFCS The Problem The Solution The Problem W D 100 1st D 100 min D 100 max D 1000 (Intercept) *** *** *** age you ** *** ** age old * *** * chil edu low edu high ** *** ** emp edu *** *** ** emp oth ** emp ret ** *** * emp self emp un fem inc core quin ctry2 ** *** inc core quin ctry3 * *** inc core quin ctry4 *** *** * inc core quin ctry5 ** *** mar div mar mar mar wid *** *** *** mult earn ** ** * w net quin ctry2 *** *** *** w net quin ctry3 *** *** *** w net quin ctry4 *** *** *** w net quin ctry5 *** *** *** Data source: HFCS (2013). ***/**/*: Significance at 1/5/10% level. Brussels, 11/03/2015 | Ertz and M¨ unnich | 2 (3) Pitfalls of regression modelling with complex survey data
The HFCS The Problem The Solution The Problem W D 100 1st D 100 min D 100 max D 1000 (Intercept) *** *** *** *** age you ** *** ** *** age old * *** * ** chil edu low edu high ** *** ** ** emp edu *** *** ** *** emp oth ** emp ret ** *** * ** emp self emp un fem inc core quin ctry2 ** *** ** inc core quin ctry3 * *** * inc core quin ctry4 *** *** * ** inc core quin ctry5 ** *** * mar div mar mar mar wid *** *** *** *** mult earn ** ** * ** w net quin ctry2 *** *** *** *** w net quin ctry3 *** *** *** *** w net quin ctry4 *** *** *** *** w net quin ctry5 *** *** *** *** Data source: HFCS (2013). ***/**/*: Significance at 1/5/10% level. Brussels, 11/03/2015 | Ertz and M¨ unnich | 2 (3) Pitfalls of regression modelling with complex survey data
The HFCS The Problem The Solution The Problem W D 100 1st D 100 min D 100 max D 1000 (Intercept) *** *** *** *** age you ** *** ** *** age old * *** * ** chil edu low edu high ** *** ** ** emp edu *** *** ** *** emp oth ** emp ret ** *** * ** emp self emp un fem inc core quin ctry2 ** *** ** inc core quin ctry3 * *** * inc core quin ctry4 *** *** * ** inc core quin ctry5 ** *** * mar div mar mar mar wid *** *** *** *** mult earn ** ** * ** w net quin ctry2 *** *** *** *** w net quin ctry3 *** *** *** *** w net quin ctry4 *** *** *** *** w net quin ctry5 *** *** *** *** Data source: HFCS (2013). ***/**/*: Significance at 1/5/10% level. Brussels, 11/03/2015 | Ertz and M¨ unnich | 2 (3) Pitfalls of regression modelling with complex survey data
The HFCS The Problem The Solution The Solution ◮ What went wrong here? ◮ How can we fix it? ◮ Your questions will be answered at our poster. ◮ We are looking forward to interesting discussions with you! Brussels, 11/03/2015 | Ertz and M¨ unnich | 3 (3) Pitfalls of regression modelling with complex survey data
Recommend
More recommend
