1 / 20

MULTICOLLINEARITY

KULIAH 12. MULTICOLLINEARITY. What’s Multicolinearity ?. Nature of the problem: X’X matrix must not be singular  why? Ada hubungan linier antar beberapa ( atau semua ) variabel bebas . Perfect: Not perfect:. ilustrasi. Penyebab.

jayme
Télécharger la présentation

MULTICOLLINEARITY

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. KULIAH 12 MULTICOLLINEARITY

  2. What’s Multicolinearity? • Nature of the problem: X’X matrix must not be singular  why? • Ada hubungan linier antarbeberapa (atausemua) variabelbebas. • Perfect: • Not perfect:

  3. ilustrasi

  4. Penyebab • Metodepengumpulan data,sampeldiambildaripopulasidgnlingkupterbatas • Keterbatasan model/populasi, ex: Y= konsumsilistrik, X1 = pendapatanruta, X2 = luasrumah • Spesifikasi model, ex: menambahkanvariabelpolinomialpada data X ygterbatas • Overdetermined model: #paramater > # obs • Common trend, ex: income, poupulation, wealth growing over time at more or less the same rate

  5. Efekthdestimasi • Perfect collinearity: no coefficient estimates , s.e. = ∞ ! Apaarti ? • High collinearity: maybe there are coefficient estimates , s.e. = very large Bgmdgn CI? • BgmkabarBLUE?

  6. Konsekuensi model ber-multicollinearity

  7. R2tinggivs t-value ApakomentarAnda ?

  8. Ceklbhljt ApakomentarAnda?

  9. Sensitivitasthdperubahan data  Estimasi parameter tidakstabil

  10. Diagnosa (ingat-2 tugasyl !) • 1. High R2 but few significant t ratios. • 2. High pair-wise correlations among regressors. (tapikdgterjadijugameskirijrendah) • 3. Examination of partial correlations. Misal:  = 1  if rij= 0.5  R2tinggitapi partial-R2rendah

  11. Diagnosa (ingat-2 tugasyl !) • 4. Auxiliary regressions. • to regress each Xi on the remaining X variables and compute the corresponding R2 (R2i ) • Fi sig  Xicollinearity with other X • Rule of thumb: R2i >R2  multicollinearity problem

  12. 5. Eigenvalues and condition index. (SAS) • -------10 ---CI----30--------- • 6. Tolerance (TOL) and variance inflation factor (VIF). severe low moderate strong

  13. VIF • r23 = koef. korelasiantara X2 dan X3 •  r23  , ,  •  r23 = 1 ?

  14. VIF • Kecepatankenaikanvar-covar variance inflation factor (VIF)

  15. VIF – RLB dgn k-variabel •  VIF   prob. multikolinierity • Rule of thumb: VIF > 10  high multicollinearity • 0 ≤ TOLj ≤ 1

  16. Remedial • Do nothing ??! • 1. Apriori information: berdasarteori or pengalamansebelumnya • didapat didapatdarihubungan

  17. Remedial • 2. Combining cross-sectional and time series data. • Time series view: Price & income sgtberkorelasi multikolinieriti •  estimate regresi (time series) • Dimana (regresicross section)

  18. Remedial • 3. Dropping a variable(s) and specification bias. • Ex: consumption = f (income, wealth) (cthsebelumnya) •  income & wealth berkorelasi  hapus wealth dari model • Tapijikateorimenyatakanbhwfungsidiatasberlaku, makamenghapus wealth dari model akanmengakibatkan bias spesifikasi. • True model: • Estimated by:  • b32 = koefregresi b3atas b2 • Jika > 0  b12 over estimate dariβ2 (bias +) • Jika< 0  b12 under estimate dariβ2 (bias -)

  19. Remedial • 4. Transformasivariabel • First differencing • Ratio transformation • Y = konsumsi, X2 = PDB, X3 = JmlPddk •  PDB & JmlPddk “grow over time”  berkorelasi • Regresi per kapitapenduduk: Be careful of new problem: serially correlated error, heteroscedasticity,

  20. Remedial • 5. Menambahjumlah data (observasi) • n = 10  • n = 40  • 6. Reducing collinearity in polynomial regressions. Transform variables in deviation form. • 7. Other methods of remedying multicollinearity, ex: factor analysis, ridge regression, principal component regression

More Related