Testing the Information Matrix Equality with Robust Estimators



Figure 10: Regression: RP-Power curves: Cauchy (a) n =10(b)
n = 20

all replications. The sample size equals 10 or 20 for the Cauchy alternative,
and 50 and 100 in the case of vertical outliers. The RP-power curves are
constructed as explained in Section 6.1. In Figure 10 the RP-power curves
are plotted for the Cauchy alternative. As in the normal model, the IM tests
with robust estimators perform better than the IM test with ML estimator,
although the difference in power is minor. In the presence of 20% vertical
outliers lying at (a)
Y =5whenn =50and(b)Y =5andY = -5 when
n = 100, the TB and MAD estimators yield much more powerful IM tests
than the ML estimator (see Figure 11). In this case, the 5%-level IM test
with the ML estimator is biased, meaning that the actual RP under the
null exceeds power. Figure 12 plots the RP-power curves for the case of
bad leverage points, with sample sizes equal to 50 and 100, and 10% and
5% bad leverage points, respectively. In both cases the IM test with robust
estimators has much more power, as expected. These simulation results
convincingly demonstrate the usefulness of robust estimators in a regression
context in connection with the IM test.

23



More intriguing information

1. Visual Perception of Humanoid Movement
2. The name is absent
3. THE CHANGING RELATIONSHIP BETWEEN FEDERAL, STATE AND LOCAL GOVERNMENTS
4. DURABLE CONSUMPTION AS A STATUS GOOD: A STUDY OF NEOCLASSICAL CASES
5. Federal Tax-Transfer Policy and Intergovernmental Pre-Commitment
6. Secondary stress in Brazilian Portuguese: the interplay between production and perception studies
7. Migrating Football Players, Transfer Fees and Migration Controls
8. Trade and Empire, 1700-1870
9. Problems of operationalizing the concept of a cost-of-living index
10. WP 1 - The first part-time economy in the world. Does it work?
11. A Hybrid Neural Network and Virtual Reality System for Spatial Language Processing
12. Yield curve analysis
13. Evidence on the Determinants of Foreign Direct Investment: The Case of Three European Regions
14. The name is absent
15. The name is absent
16. Strategic Planning on the Local Level As a Factor of Rural Development in the Republic of Serbia
17. Governance Control Mechanisms in Portuguese Agricultural Credit Cooperatives
18. Rent-Seeking in Noxious Weed Regulations: Evidence from US States
19. Towards Learning Affective Body Gesture
20. Searching Threshold Inflation for India