metaLik: Likelihood inference in meta-analysis and meta-regression models
metaLik is a package for the R statistical computing environment, which implements first- and higher-order likelihood inference (Skovgaard's statistic) in meta-analysis and meta-regression problems. The package is based on the methodology developed in Guolo (2012). Benefits from higher-order asymptotics arise in the not so unusual case of small number of meta-analysis studies. In such a case, ordinary first-order likelihood inference may be very imprecise, thus resulting in wrong inferential conclusions. The package authors are Annamaria Guolo and Cristiano Varin. A detailed description of metaLik can be found in Guolo and Varin (2012).
The metaLik software is made available free, under GPL-3. It comes with ABSOLUTELY NO WARRANTY.
Installing metaLik from CRAN
R> install.packages( "metaLik" )
R> library( metaLik )
Meta Regression of Vaccine Data
Next lines illustrates metaLik analysis of thirteen clinical studies evaluating the efficacy of the BCG vaccine for the prevention of tuberculosis. The aim is to evaluate a potential latitude effect on vaccination efficacy.
R> data( vaccine )
R> m <- metaLik( y ~ latitude, data = vaccine, sigma2 = sigma2 )
R> summary( m )
Likelihood inference in random-effects meta-analysis models
Call:
metaLik(formula = y ~ latitude, data = vaccine, sigma2 = sigma2)
Estimated heterogeneity parameter tau2 = 0.1676
Test for heterogeneity Q = 63.64 (pval < 0.0001)
Fixed-effects:
estimate std.err. signed logLRT p-value Skovgaard p-value
(Intercept) -0.3050 0.2241 -1.3378 0.181 -1.2245 0.2208
latitude -0.0154 0.0064 -2.1203 0.034 -1.8164 0.0693
Log-likelihood: 1.1212
First-order inference suggests a significant effect of latitude, while higher-order asymptotics is doubtful about such an effect. Plotting profile log-likelihood is useful for visualization.
R> profile( m )
metaLik is a package for the R statistical computing environment, which implements first- and higher-order likelihood inference (Skovgaard's statistic) in meta-analysis and meta-regression problems. The package is based on the methodology developed in Guolo (2012). Benefits from higher-order asymptotics arise in the not so unusual case of small number of meta-analysis studies. In such a case, ordinary first-order likelihood inference may be very imprecise, thus resulting in wrong inferential conclusions. The package authors are Annamaria Guolo and Cristiano Varin. A detailed description of metaLik can be found in Guolo and Varin (2012).
The metaLik software is made available free, under GPL-3. It comes with ABSOLUTELY NO WARRANTY.
Installing metaLik from CRAN
R> install.packages( "metaLik" )
R> library( metaLik )
Meta Regression of Vaccine Data
Next lines illustrates metaLik analysis of thirteen clinical studies evaluating the efficacy of the BCG vaccine for the prevention of tuberculosis. The aim is to evaluate a potential latitude effect on vaccination efficacy.
R> data( vaccine )
R> m <- metaLik( y ~ latitude, data = vaccine, sigma2 = sigma2 )
R> summary( m )
Likelihood inference in random-effects meta-analysis models
Call:
metaLik(formula = y ~ latitude, data = vaccine, sigma2 = sigma2)
Estimated heterogeneity parameter tau2 = 0.1676
Test for heterogeneity Q = 63.64 (pval < 0.0001)
Fixed-effects:
estimate std.err. signed logLRT p-value Skovgaard p-value
(Intercept) -0.3050 0.2241 -1.3378 0.181 -1.2245 0.2208
latitude -0.0154 0.0064 -2.1203 0.034 -1.8164 0.0693
Log-likelihood: 1.1212
First-order inference suggests a significant effect of latitude, while higher-order asymptotics is doubtful about such an effect. Plotting profile log-likelihood is useful for visualization.
R> profile( m )
References
Guolo A. (2012). Higher-order likelihood inference in meta-analysis and meta-regression. Statistics in Medicine, 31 (4), 313–327.
Guolo A. and Varin C. (2012). The R package metaLik for likelihood inference in meta-analysis. Journal of Statistical Software, 50 (7), 1–14.
Skovgaard I. M. (1996). An explicit large-deviation approximation to one-parameter tests. Bernoulli, 2, 145–165.
Guolo A. (2012). Higher-order likelihood inference in meta-analysis and meta-regression. Statistics in Medicine, 31 (4), 313–327.
Guolo A. and Varin C. (2012). The R package metaLik for likelihood inference in meta-analysis. Journal of Statistical Software, 50 (7), 1–14.
Skovgaard I. M. (1996). An explicit large-deviation approximation to one-parameter tests. Bernoulli, 2, 145–165.