* KULIAH 11
*Heteroskedasticity*Serial correlation
*Multicollinerity
*Normality
*Omitted variables
*What’s Heteroskedasticity?
*Varians residual tdk konstan
Prototype
*Penyebab*Error learning misal: belajar mengetik
*Sampel yang beragam rumahtangga dgn pendptn, perusahaan berbagai level
*Adanya outlier
*Omitting variables
*Sebaran data tidak normal
*incorrect data transformation (e.g., ratio or first difference transformations) and
*incorrect functional form (e.g., linear versus log–linear models)
* lebih sering terjadi pada data cross section
*Efek thd estimasi
*BLUE?
*Linear Unbiased but not efficient LU
Homoscedastic?
Which is the Homoscedast
ic?
*KOnsekuensi*Bagaimana estimasi yg diperoleh terkait varians yg tidak
konstan?
*- Signifikansi ?
*- CI ?
* misleading …
*Mendeteksi heteroskedasticity *Nature of problem (functional form review )
*Periksa Grafik residual
*Tes statistik
*Tes Statistik*Bahwa residual berkorelasi dengan varians
*Park Test
* signifikan residuals are heteroskedastic
* weakness: may not satisfy the OLS assumptions and may itself be heteroscedastic
*Glejser Test
* weakness: the error term vi has some
problems in that its expected value is nonzero,
it is serially correlated and ironically it is
heteroscedastic, some models are non linear.
*Ex: Park & Glejser test
H0: residuals are homoskedasticH1: residuals are heteroskedastic
*Goldfeld-Quandt Test: the heteroscedastic variance, σ2i , is
positively related to one of the explanatory variables in the regression model, ex:
* σ2i would be larger, the larger the values of Xi
*Weakness:
*- depend on which c is arbitrary,
*- for X > 1 Var, which X is correct to be ordered?
*Ex:*Y = Income,
*X = Consumption,
*n = 30,
*c = 4
*Ex: *Y = Income, X = Consumption, n = 30, c = 4
*Breusch–Pagan–Godfrey Test
*Weakness: - large sample needed for small sample, depend much on normality assumption
Ex:
So, H0: residuals are Homoskedastic
ESS = SSR
*Ex:
𝜒❑2 (1,5% )=3,8414❑
*White’s General Heteroscedasticity Test.
*Weakness: more variables will consume more df.
H0: residuals are homoskedastic
Or H0:
, df = # parameter -1
*Koenker–Bassett (KB) test.
*H0: residuals are homoskedastic
*Or H0:
*Tes hipotesis using t-test
Obtain residual, then estimate
*Other tests…..*Find other references…
*Remedial
Perhatikan 1 &
2
Reparameterize before analize !
Reparameterize before analize !
*Practically, run OLS first, then run:
* consistent estimator large sample needed
* measure the elasticity
*Other Remedial Procedure
*Run the following (weighted) regression:
*Compare with the unweighted
Apa perbedaan
kedua model ini?
*White suggests:
*For RLB:
*Important notes
*Tugas Bonus*Pelajari Gujarati, Basic
Econometrics, 14th edition,
*Ch. 11, section 11.7