The Exact Distribution of the TSLS Estimator for a NonGaussian Just-Identified Linear Structural Equation by Giovanni Forchini

September 13, 2006

Supplementary results Expressing φw2 ( t1 ) as c2 Φ c ( t1 | w22 ) φm ( w22 ) 2

Notice that

φw ( t1 ) = c2 2 2

=



0< q


φ (q + w22 )q −1/ 2 dq × c2φm ( w22 ) 2 φm ( w2 )



pdf ⎛⎜⎝ t1 | w22 ⎞⎟⎠ dq × c2φm ( w22 )

0< q
=

φ (q + w22 )q −1/ 2 dq

0< q
= Φ c ( t1 | w22 ) c2φm ( w22 ) .

Detailed proof of Proposition 1 The starting point is the density of βˆ when Π = 0 ,

( )

−1/ 2 pdf βˆ = c2 det ( Ω )



ˆ −∞<Π<+∞

( )

( )

ˆ. ˆ φ ⎛⎜⎜ a βˆ Π ˆ 2 ⎞⎟⎟ d Π Π ⎝

( )



ˆ to Π ˆ = x / a βˆ . The Jacobian is 1/ a βˆ , so We transform Π

1

( )

−1/ 2 pdf βˆ = c2 det ( Ω )



−∞< x <+∞

= c2 det ( Ω )

( )

−1/ 2

a βˆ



( )

( )

( )

2⎞

( )

⎜ ⎟ x / a βˆ φ ⎜⎜ a βˆ ⎡ x / a βˆ ⎤ ⎟⎟ / a βˆ dx ⎢ ⎥ ⎜ ⎣ ⎦ ⎟⎠ ⎝

−1



−∞< x <+∞

x φ ⎛⎜⎝ x 2 ⎞⎟⎠ dx

Now we observe that the integrand is symmetric around zero so that

( )

( ) ∫

−1 / 2 pdf βˆ = 2c2 det ( Ω ) a βˆ

−1

0 < x <+∞

x φ ⎛⎜⎝ x 2 ⎞⎟⎠ dx ,

( )

and set x = t . The Jacobian of this transformation is 1/ 2 t , so

( )

( ) ∫ −1

−1 / 2 pdf βˆ = c2 det ( Ω ) a βˆ

φ ( t ) dx .

0 < t <+∞

Equation (13) follows from the fact that (10) integrates to one so that



π c2

φ ( t ) dx = 1 .

0 < t <+∞

Detailed proof of Proposition 2 ˆ as follows We start from equation (12) In the general case, we transform Π

( )

ˆ = a βˆ Π

−1/ 2

(

( )( )

s + b βˆ a βˆ

−1/ 2

( )

) . The Jacobian is 1/

( ) ∫

−1/ 2 pdf βˆ = c2 det ( Ω ) a βˆ

−1

( )( )

s + b βˆ a βˆ

−∞< s <+∞

( )

a βˆ , and the density of βˆ is −1/ 2

(

( ) ( )) 2

φ s 2 + c − b βˆ /a βˆ ds .

Next we transform s using polar coordinates, s = ht 1/ 2 , where h 2 = 1 and t > 0 (the Jacobian is (1/ 2)t −1/ 2 ) we have:

( )

( ) ∫

−1/ 2 pdf βˆ = c2 det ( Ω ) a βˆ

(A.1) (1/ 2)

−1

0


( )( )

ht1/ 2 + b βˆ a βˆ

h =1

(

( ) ( )) 2

φ t + c − b βˆ /a βˆ t −1/ 2 −1 / 2

( dh ) dt.

2

The integral over h 2 = 1 can be evaluated explicitly. Since h can take on only the values ±1 , the integral is in fact a sum:

2

( )( )

1 ht1/ 2 + b βˆ a βˆ ∫ 2 h2 =1

−1/ 2

( dh ) =

( )( )

1⎡ ( +1) t1/ 2 + b βˆ a βˆ ⎢ 2⎣

−1 / 2

( )( )

−1/ 2

( ) ( )

−1 / 2

+ ( −1) t1/ 2 + b βˆ a βˆ

( ) ( )

2 The term on the right-hand-side equals t 1/ 2 if t > b βˆ /a βˆ

and b βˆ a βˆ

⎤ ⎥⎦ .

if

( ) ( ) 2

0 < t < b βˆ /a βˆ . Inserting the last display into equation (A.1) we obtain

( )

( ) ( )

−1/ 2 pdf βˆ = c2 det ( Ω ) b βˆ a βˆ

−3 / 2



(

( ) ( ) 2

0
(A.2) +c2 det ( Ω )

−1/ 2

( )

a βˆ

−1





0< t < b ( βˆ )2 /a ( βˆ )

(

2

(

( ) ( )) 2

φ t + c − b βˆ /a βˆ dt.

b ( βˆ ) 2 /a ( βˆ )
Notice that using (11) the integral in the first line is c2

( ) ( ))

φ t + c − b βˆ /a βˆ t −1/ 2 dt

( ) ( )) 2

φ t + c − b βˆ /a βˆ t −1/ 2 dt = φ

b( βˆ ) /a ( βˆ ) ) . ( ) ( )( 2

2

c −b βˆ /a βˆ

( ) ( )

2 To evaluate the integral in the second line of (A.2), let t = z − ⎡c − b βˆ /a βˆ ⎤ , (the ⎢⎣ ⎥⎦

Jacobian is 1) so that

π c2



φ ( z ) dz = 1 − Φ ( c ) .

c < z <+∞

3

The Exact Distribution of the TSLS Estimator for a Non ...

Sep 13, 2006 - q t m m. q t c m. t c. q w q dq. q w q dq c w w pdf t w dq c w. t w c w φ φ φ φ φ φ φ. −. < <. −. < <. ⎛. ⎞. ⎜. ⎟. ⎝. ⎠. < <. = +. +. = ×. = ×. = Φ. ∫. ∫.

49KB Sizes 1 Downloads 124 Views

Recommend Documents

The Exact Distribution of the Hansen-Jagannathan Bound
confidence intervals for the constrained HJ bound (dotted lines) using the methodology described later in the paper. The confidence intervals in Figure 2 are quite wide, indicating that there is substantial uncertainty about the exact location of the

The Size Distribution of Exporting and Non-exporting ...
distribution of exporting firms among Chinese provinces, in varying degrees, are less than 1 ... in 2007. This relationship between the power law exponents of exporting and non-exporting firms is in .... details are in an online appendix Table 5.

Properties of the Maximum q-Likelihood Estimator for ...
variables are discussed both by analytical methods and simulations. Keywords ..... It has been shown that the estimator proposed by Shioya is robust under data.

A non-cooperative interpretation of the reverse Talmud rule for ...
Click here to download Manuscript: The RT rule.pdf Click here to view linked References. 1. 2. 3. 4 ... es Equal treatment of equals, Resource monotonicity, and.

pdf-1442\electric-power-distribution-systems-a-non-technical-guide ...
... more apps... Try one of the apps below to open or edit this item. pdf-1442\electric-power-distribution-systems-a-non-technical-guide-by-wayne-beaty.pdf.

KAAD-BR Distribution and Posting of Non-Curricular Material in ...
There will be no charge for putting an item in the existing employee. newsletter, the intercom. A group can also request that material be posted in staff. lounges. Page 3 of 4. KAAD-BR Distribution and Posting of Non-Curricular Material in School.pdf

A Fast and Exact Algorithm for the Median of Three ...
School of Computer and .... The vertices of simple adequate subgraphs of rank 3 have degrees either 2 or 3. ... subgraph, there exists a vertex x with degree 2.

A Fast and Exact Algorithm for the Median of Three ...
ulated data show dramatic speedup so that many instances can be solved very .... for the median of three problem is 3-regular, the vertex degrees of its induced.

A fast and exact algorithm for the median of three ...
can be used to decompose multiple breakpoint graphs, achieving a dra- matic speedup in solving the ... red matching, we mean the matching consisting of all the red edges. The size of the breakpoint ...... TR CRM-2579. Centre de recherches ...

Exact confidence intervals for the Hurst parameter of a ... - Project Euclid
sian and non-Gaussian stochastic processes, is crucial in applications, ranging ..... of Hermite power variations of fractional Brownian motion. Electron. Comm.

Exact and heuristic methods for the selective ...
Apr 10, 2008 - so as to maximize the system reliability, within a time window of a limited ...... ISA – The Instrumentation, System, and Automation Society, 2004.

New exact algorithms for the 2-constraint satisfaction ...
bound on the size of a vertex separator for graphs in terms of the average degree of the graph. We then design a simple algorithm solving MAX-2-CSP in time O∗(2cdn), cd = 1 − 2α ln d d for some α < 1 and d = o(n). Keywords: exact exponential ti

A Sparse Structured Shrinkage Estimator for ...
Jan 11, 2011 - on model selection consistency and estimation bounds are derived. ..... The gradient and Jacobian of the objective function in (7) are, respectively,. Grg ...... the SSS procedure can indeed recover the motifs related to the cell ...

A Least#Squares Estimator for Monotone Index Models
condition which is weaker than what is required for consistency of Hangs MRC. The main idea behind the minimum distance criterion is as follows. When one ...

A Sparse Structured Shrinkage Estimator for ...
Jan 11, 2011 - for high-dimensional nonparametric varying-coefficient models and ... University of Pennsylvania School of Medicine, Blockley Hall, 423 .... the Appendix, available as online supplemental materials. 3 ...... Discovery and Genome-Wide E

High-level Distribution for the Rapid Production of ...
Investigating the potential advantages of the high-level Erlang tech- nology shows that ..... feature of many wireless communication systems. Managing the call ...

New Exact and Approximation Algorithms for the ... - Research at Google
We show that T-star packings are reducible to network flows, hence the above problem is solvable in O(m .... T and add to P a copy of K1,t, where v is its center and u1,...,ut are the leafs. Repeat the .... Call an arc (u, v) in T even. (respectively

Simple and Exact Closed-Form Expressions for the ... - IEEE Xplore
results are derived for Nakagami-m fading. The results derived in this paper thus allow for lock detector threshold values, lock probabilities, and false alarm rates ...

High-level Distribution for the Rapid Production of Robust Telecoms ...
time performance is achieved, e.g. three times faster than the C++ imple- ..... standard packet data in GSM systems [9], and the Intelligent Network Service.