Description Usage Arguments Details Value See Also Examples
This function enables the visual analysis of the confidence sets generated by the likelihood function by creating a sample of targeted parameters from a confidence distribution.
1 | independent(model, y, fit, cov, B, max_B)
|
model |
model of the targeted analysis. Current available models are: Gaussian Regression, Poisson Regression, Logistic Regression, Negative Binomial Regression and Gamma Regression (both inverse and log). |
y |
response variables |
fit |
basic statistics after fitting the model by R. Most of the model fits are available in GLM family. If the model cannot be fitted automatically, user must provide at least the list of response variable, design(X) matrix, maximum likelihood estimators, maximum log-likelihood values, and the inference function. |
cov |
the covariance matrix of the model. System will use the default covariance matrix, unless given by user. |
B |
number of rays to be generated. |
max_B |
maximum number of rays to be generated for each parameter. Default value is 1000. |
The method generates a sample of parameters from a confidence distribution, which is designed so that its probabilities on the parameters are equal to the asymptotic coverage probabilities of the targeted confidence sets.
For certain models, the covariance matrix, cov
includes the mean square error of the model. For instance, Gaussian Regression. These models will have different covariance matrices, which contains only the MLE.
MLE |
the maximum likelihood estimators of the model |
max_loglik |
the maximum value of the log-likelihood function |
diagnosis |
how many times does the ray generated intersects the boundary. They are categorized as “0”, “1”, “2” and “>2”. Each confidence level is displayed separately. |
independent.sample |
the generated samples by boundary sampling method. |
ray.z |
the randomly generated multivariate normal values. |
numWald.interval |
numerical Wald intervals for each maximum likelihood estimator. |
simWald.interval |
simulated Wald intervals for each maximum likelihood estimator. |
boundary
for boundary sampling method.
Mgaussian
for gaussian regression.
Mpoisson
for poisson/log-linear regression.
Mlogistic
for logistic regression.
Mgamma
for gamma regression.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 | library(MASS)
data(data.poisson)
Device <- data.poisson$Device
DataPlan <- data.poisson$DataPlan
y <- data.poisson$y
fit <- glm(y ~ (Device+DataPlan), family=poisson)
out_i <- independent("Mpoisson",y,fit,B=1000) # Simulation with desired number of rays
out_i <- independent("Mpoisson",y,fit) # Simulation without defining B
out_i$diag # out_i$diag is equivalent to out_i$diagnosis
out_i$ind[100:140,] # out_i$ind is equivalent to out_i$independent.sample
out_i$num #out_i$num is equivalent to out_i$numWald.interval
out_i$sim #out_i$sim is equivalent to out_i$simWald.interval
## Visualization
par(mfrow=c(2,2))
plot(out_i$ind[,7]~out_i$ind[,6],
xlab=expression(beta[1]),ylab=expression(beta[2]),cex=0.5)
points(out_i$MLE[2],out_i$MLE[3],pch=16,col="red",cex=1.5)
plot(out_i$ind[,6]~out_i$ind[,5],
xlab=expression(beta[0]),ylab=expression(beta[1]),cex=0.5)
points(out_i$MLE[1],out_i$MLE[2],pch=16,col="red",cex=1.5)
plot(out_i$ind[,7]~out_i$ind[,5],
xlab=expression(beta[0]),ylab=expression(beta[2]),cex=0.5)
points(out_i$MLE[1],out_i$MLE[3],pch=16,col="red",cex=1.5)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.