xtlogit - Stata
xtlogit - Stata
xtlogit - Stata
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Methods and formulas<br />
<strong>xtlogit</strong> — Fixed-effects, random-effects, and population-averaged logit models 15<br />
<strong>xtlogit</strong> reports the population-averaged results obtained using xtgee, family(binomial)<br />
link(logit) to obtain estimates. The fixed-effects results are obtained using clogit. See [XT] xtgee<br />
and [R] clogit for details on the methods and formulas.<br />
If we assume a normal distribution, N(0, σ 2 ν), for the random effects ν i ,<br />
Pr(y i1 , . . . , y ini |x i1 , . . . , x ini ) =<br />
∫ ∞<br />
−∞<br />
e −ν2 i /2σ2 ν<br />
√<br />
2πσν<br />
}<br />
∏<br />
F (y it , x it β + ν i ) dν i<br />
{<br />
ni<br />
t=1<br />
where<br />
⎧<br />
1<br />
⎪⎨<br />
1 + exp(−z)<br />
F (y, z) =<br />
1 ⎪⎩<br />
1 + exp(z)<br />
if y ≠ 0<br />
otherwise<br />
The panel-level likelihood l i is given by<br />
l i =<br />
∫ ∞<br />
−∞<br />
e −ν2 i /2σ2 ν<br />
√<br />
2πσν<br />
{<br />
ni<br />
}<br />
∏<br />
F (y it , x it β + ν i ) dν i<br />
t=1<br />
≡<br />
∫ ∞<br />
−∞<br />
g(y it , x it , ν i )dν i<br />
This integral can be approximated with M-point Gauss–Hermite quadrature<br />
This is equivalent to<br />
∫ ∞<br />
−∞<br />
∫ ∞<br />
−∞<br />
e −x2 h(x)dx ≈<br />
f(x)dx ≈<br />
M∑<br />
wmh(a ∗ ∗ m)<br />
m=1<br />
M∑<br />
wm ∗ exp { (a ∗ m) 2} f(a ∗ m)<br />
m=1<br />
where the wm ∗ denote the quadrature weights and the a∗ m denote the quadrature abscissas. The log<br />
likelihood, L, is the sum of the logs of the panel-level likelihoods l i .<br />
The default approximation of the log likelihood is by adaptive Gauss–Hermite quadrature, which<br />
approximates the panel-level likelihood with<br />
l i ≈ √ 2̂σ i<br />
M ∑<br />
m=1<br />
w ∗ m exp { (a ∗ m) 2} g(y it , x it , √ 2̂σ i a ∗ m + ̂µ i )<br />
where ̂σ i and ̂µ i are the adaptive parameters for panel i. Therefore, with the definition of g(y it , x it , ν i ),<br />
the total log likelihood is approximated by