A COMPANION TO Theoretical Econometrics
Models and Their Specification
Suppose the focus of the analysis is to consider the behavior of the n x 1 vector of random variables wt = (w1t, w2t,..., wnt)' observed over the period t = 1, 2,..., T. A model of wt, indexed by Щ-, is defined by the joint probability distribution function (pdf) of the observations
[11] + exp[(Zfe - ZitYy2 + Y 1(yis-1 - yi, t+1) + Y 1(yis+1 - yi, t-1)]1(t - s ^ 3)
(16.30)
[12] are unemployed in January and remain unemployed in December too;
[13] are unemployed in January and find a job before December.
[14] MC tests based on pivotal statistics: an exact randomized test procedure;
• MC tests in the presence of nuisance parameters:
(a) local MC p-value,
(b) bounds MC p-value,
(c) maximized MC p-value;
• MC tests versus the bootstrap:
(a) fundamental differences/similarities,
(b) the number of simulated samples: theory and guidelines;
[15] Let z = (X1,..., Xm, Y1,..., Yn) and s = m Xf 1X - n Щ=1 Y.
• Obtain all possible Q = (n + m)! permutations of z, z(1),..., z(Q), and calculate the associated "permuted analogs" of s
[16] From the observed data, compute:
(a) the test statistic S0, and
(b) a restricted consistent estimator P0n of 0.
[17] Compute 10 and 1, the restricted and unrestricted SURE (iterative) MLE.
• Compute 1 „ as the unconstrained (OLS) estimate of 1 in the "nesting" MLR model.
• Compute Л* = 110|/| 11 and LR* = n 1п(Л*).
• Draw 99 realizations from a multivariate (n, 3, I) normal distribution: U(1), U(2), ..., U(p) and store.
• Consider the linear constraints
t-S
t=1
