Displaying 2 results from an estimated 2 matches for "sumlog".
Did you mean:
sublog
2009 Jan 29
1
Arima_Like() and NaN - a (possible) problem, a patch, and RFC
...e code (file src/arima.c of the stats package) and
noticed that this second element is a sum of logarithmic terms, computed
through the following snippet of code:
gain = M[0];
for (j = 0; j < d; j++) gain += delta[j] * M[r + j];
if(gain < 1e4) {
nu++;
ssq += resid * resid / gain;
sumlog += log(gain);
}
Here, sumlog is the second element of the resulting vector. However, the
"if(gain < 1e4) {" check does not explicitly check against negative
values of the gain variable. Indeed, whenever the gain variable assumes
a negative value, the statement "sumlog += log(gain...
2006 Jun 02
2
Problem with mle
..., 288, 260, 153, 116, 57, 33, 40, 44, 36,
24, 35, 23, 36, 25, 35, 30)
X <- c(1.74924, 3.49848, 5.24772, 6.99696, 8.74620, 17.49240, 26.23860,
34.98480, 43.73100, 52.47720, 61.22340, 69.96960, 71.71880, 73.46810,
75.21730, 76.96650, 78.71580, 80.46500 )
#Sum of log of values in vector
SumLog <- function (x)
{
sum(log(x))
}
#-ln(L) with Poisson Likelihood estimator
NeglogPoisL <- function (obs, est)
{
sel <- est != 0
SumLogObs <- SumLog(obs[sel])
-sum( obs[sel] * log(est[sel]) - est[sel]- SumLogObs )
}
#-ln(L) for Negative Exponential
NENeglogPoisL <- function(a=0...