tabmeta: Publication-Ready Meta-Analysis in One Command
In R4VN: Health Data Analysis and Publication-Ready Reporting
tabmeta R Documentation
Publication-Ready Meta-Analysis in One Command
Description
Fits fixed- and/or random-effects meta-analysis and returns a complete,
publication-ready report. tabmeta() accepts a binary 2-by-2 table
(a, b, c, d), event/total data, continuous summaries, rates,
correlations, or study-level effect estimates with standard errors or
confidence intervals. The default profile="auto" adds prediction,
few-study inference, subgroup/moderator results when requested,
small-study-effect diagnostics, sensitivity analyses, and figures.
Usage
tabmeta(
data = NULL,
study,
effect = NULL,
se = NULL,
lower = NULL,
upper = NULL,
a = NULL,
b = NULL,
c = NULL,
d = NULL,
event1 = NULL,
n1 = NULL,
event0 = NULL,
n0 = NULL,
mean1 = NULL,
sd1 = NULL,
mean0 = NULL,
sd0 = NULL,
event = NULL,
n = NULL,
time = NULL,
time1 = NULL,
time0 = NULL,
or = FALSE,
rr = FALSE,
rd = FALSE,
hr = FALSE,
irr = FALSE,
md = FALSE,
smd = FALSE,
prop = FALSE,
rate = FALSE,
cor = FALSE,
fixed = FALSE,
random = TRUE,
method = "REML",
hk = NULL,
small = c("auto", "adhoc", "knha", "t", "z"),
prediction = NULL,
by = NULL,
subgroup = NULL,
reg = NULL,
moderator = NULL,
bias = NULL,
bias_methods = "auto",
leaveout = NULL,
influence = NULL,
cumulative = NULL,
transform = NULL,
cc = 0.5,
zero = c("keep", "exclude"),
ci = 0.95,
digit = 2,
p_digit = 3,
profile = c("auto", "brief", "full", "custom"),
full = NULL,
plot = NULL,
plot_display = "forest",
plot_args = list(),
report = FALSE,
interpretation = FALSE,
title = NULL,
show = TRUE,
export = NULL,
file = NULL,
open = FALSE,
strict = FALSE
)
Arguments
data
Optional data frame. When omitted, active R4VN data are used.
study
Study label variable.
effect
Generic study-level effect estimate on the natural scale.
se
Standard error on the analysis scale. For ratio measures this is
the standard error of the log effect.
lower, upper
Lower and upper confidence limits for effect.
a, b, c, d
Binary 2-by-2 cells: events and non-events in group 1
(a, b) and group 0 (c, d). Supplying these four arguments is
equivalent to event1=a, n1=a+b, event0=c, n0=c+d.
event1, n1, event0, n0
Events and total sample sizes in groups 1 and 0.
mean1, sd1, mean0, sd0
Group means and standard deviations.
event, n, time
Single-group events, sample size, and person-time.
time1, time0
Person-time in groups 1 and 0 for incidence-rate ratios.
or, rr, rd, hr, irr, md, smd, prop, rate, cor
Logical effect selectors.
With raw binary input, OR is inferred if none is selected; with continuous
summaries, MD is inferred. Set a selector to request another measure.
fixed
Fit a fixed-effect model in addition to, or instead of, the
random-effects model.
random
Fit a random-effects model; default TRUE.
method
Random-effects tau-squared estimator; default "REML".
hk
Backward-compatible logical shortcut. TRUE is
small="knha"; FALSE is small="z". Prefer small in new code.
small
Random-effects inference: "auto", "adhoc", "knha",
"t", or "z". "auto" uses modified Hartung-Knapp ("adhoc")
when there are at most 10 studies and the normal approximation otherwise.
prediction
Add a prediction interval. NULL uses the profile default.
by, subgroup
One or more subgroup variables. Use one unquoted variable,
a character vector, or vars(region, design). subgroup is a readable
alias for by; both may be combined. Factor levels and value labels from
labelled data are used in tables, interpretation, and subgroup figures.
For every estimable subgroup, the result includes its pooled effect and
confidence interval, effect test/df/p-value, prediction interval,
Cochran's Q/df/p-value, I-squared and tau-squared with confidence
intervals when estimable. The publication table is transposed: statistics
are rows and subgroup labels are columns. Multiple subgroup variables
produce one transposed table per variable.
reg, moderator
One or more meta-regression moderators created with
vars() or supplied as character names. moderator is an alias for
reg. The output includes the multivariable model and a univariable
moderator screen.
bias
Assess small-study effects/publication bias. NULL uses the
profile default.
bias_methods
One or more of "egger", "begg", "trimfill",
"failsafe", or "selection". "auto" runs Egger, Begg, and
trim-and-fill; the "full" profile also requests fail-safe N and a
selection model. Methods are used as sensitivity diagnostics, not as proof
that publication bias is or is not present.
leaveout
Perform leave-one-out sensitivity analysis. NULL uses the
profile default.
influence
Perform influence diagnostics. NULL uses the profile
default.
cumulative
Optional variable defining the ordering for cumulative
meta-analysis, usually publication year.
transform
Single-proportion transformation: "logit", "arcsine",
"ft", or "none".
cc
Continuity correction for zero events; default 0.5.
zero
Handling of double-zero binary studies: "keep" or "exclude".
ci
Confidence level as a proportion.
digit
Number of decimals for effect estimates and confidence limits,
including subgroup and meta-regression estimates. Default 2.
p_digit
Number of decimals for p-values.
profile
Analysis profile. "auto" (default) creates a complete
appropriate analysis; "brief" creates the main model and forest plot;
"full" also requests extended bias diagnostics and radial/L'Abbe plots;
"custom" turns optional modules off unless explicitly requested.
full
Backward-compatible profile shortcut: TRUE selects "full"
and FALSE selects "custom".
plot
Create publication-ready plots. NULL uses the profile default.
With subgroup analysis, the HTML Viewer includes the overall forest plot
and one clearly titled forest plot for every estimable subgroup level.
plot_display
Plot type(s) also drawn in the interactive R/RStudio Plot
pane when plot=TRUE. Default "forest". Use a character vector for Plot
history, "all" for every available figure, or "none" to suppress Plot
pane drawing while retaining figures in the HTML Viewer. When subgroup
analysis is present, selecting "forest" adds the overall forest plot and
every subgroup forest plot to Plot history; use Previous/Next to review.
plot_args
Named list of plot options. Supply common options directly,
or nested lists such as list(all=list(color="navy"), forest=list(xlim=c(0.2, 2))).
report
Add manuscript-style results text.
interpretation
Add a detailed, sectioned interpretation covering the
pooled effect, heterogeneity, prediction interval, subgroup differences,
moderators, small-study effects, few-study inference, influence,
leave-one-out robustness, and cumulative evidence whenever available.
Default is FALSE.
title
Report title.
show
Open the complete HTML report; default TRUE.
export
Optional direct export format(s): "html", "docx",
"xlsx", "pdf", or "png".
file
Export file or base path. The extension may determine the format.
open
Open the exported file.
strict
Stop when an optional diagnostic or figure cannot be created.
The default FALSE retains the main analysis and issues a warning/note.
Value
Invisibly returns an object of class r4vn_meta and r4vn_tab.
Stable publication components are available in $estimates, $tests,
$diagnostics, $models, $tables, and $metadata. The table
$tables$Statistical_tests gives the pooled-effect and Cochran's Q tests.
Its significance and conclusion columns are added only when
interpretation=TRUE. $tables$Subgroup is the transposed publication
table for the first subgroup variable, $subgroup_tables contains every
transposed subgroup table, and $subgroup_long retains tidy long output.
See Also
tabexport, vars
Other R4VN tables:
tab(),
tabexport(),
tabforest(),
tablong(),
tabmulti(),
tabscale(),
tabscore(),
tabsurvey(),
vars()
Examples
if (requireNamespace("metafor", quietly = TRUE)) {
dat <- read.csv(
system.file("extdata", "meta_example.csv", package = "R4VN")
)
# 1. Binary outcome from a, b, c, d. OR is inferred automatically.
dat$non_event_treat <- dat$n_treat - dat$event_treat
dat$non_event_control <- dat$n_control - dat$event_control
m_abcd <- tabmeta(
data = dat, study = study,
a = event_treat, b = non_event_treat,
c = event_control, d = non_event_control,
profile = "custom", plot = FALSE, show = FALSE
)
m_abcd$estimates$overall
m_abcd$tables$Binary_2x2
m_abcd$tables$Statistical_tests
m_abcd$tests$overall
m_abcd$tests$heterogeneity
plot(
m_abcd, type = "forest", show_abcd = TRUE,
abcd_titles = c("Events T", "No event T", "Events C", "No event C")
)
# 2. Equivalent event/total syntax; request RR or RD with rr/rd=TRUE.
m_or <- tabmeta(
data = dat, study = study,
event1 = event_treat, n1 = n_treat,
event0 = event_control, n0 = n_control,
or = TRUE, profile = "custom", plot = FALSE, show = FALSE
)
# 3. Complete automatic analysis: report, prediction, diagnostics, plots.
m_all <- tabmeta(
data = dat, study = study,
a = event_treat, b = non_event_treat,
c = event_control, d = non_event_control,
show = FALSE
)
# 4. Subgroup analysis: one or several subgroup variables.
m_sub <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
subgroup = region,
profile = "custom", plot = FALSE, show = FALSE
)
m_sub$tables$Subgroup
m_sub$subgroup_long
m_sub$tables$Subgroup_test
# Each subgroup row includes its effect test and heterogeneity test.
names(m_sub$tables$Subgroup)
# If the subgroup variable carries value labels, R4VN prints those labels
# instead of numeric codes. With plot=TRUE, both Viewer and Plot history
# contain the overall plot and one clearly titled plot per subgroup level.
dat$risk_group <- rep(c(1, 2), length.out = nrow(dat))
attr(dat$risk_group, "label") <- "Baseline risk"
attr(dat$risk_group, "labels") <- c("Lower risk" = 1, "Higher risk" = 2)
m_labelled <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
subgroup = risk_group, profile = "custom",
plot = TRUE, plot_display = "forest", show = FALSE
)
m_labelled$plot_titles
# Multiple subgroup analyses can be requested together.
dat$period <- ifelse(dat$year < median(dat$year), "Earlier", "Later")
m_sub2 <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
subgroup = vars(region, period),
profile = "custom", plot = FALSE, show = FALSE
)
# 5. Meta-regression with numeric and categorical moderators.
attr(dat$year, "label") <- "Publication year"
attr(dat$region, "label") <- "Geographic region"
m_reg <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
moderator = vars(c.year, region),
profile = "custom", plot = FALSE, show = FALSE
)
m_reg$tables$Meta_regression
m_reg$tables$Moderator_univariable
# Intercept is written in full; moderator labels are used when available.
# Set digit=4, for example, when four decimal places are required.
# 6. Publication-bias and small-study-effect sensitivity analyses.
m_bias <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
bias = TRUE,
bias_methods = c("egger", "begg", "trimfill", "failsafe"),
profile = "custom", plot = FALSE, show = FALSE
)
m_bias$tables$Publication_bias
m_bias$tables$Publication_bias_adjusted
# 7. Few-study inference. auto uses modified Hartung-Knapp at <=10 studies.
m_few <- tabmeta(
data = dat[1:8, ], study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
small = "auto", profile = "custom", plot = FALSE, show = FALSE
)
m_few$tables$Small_sample_inference
# 8. Generic hazard ratios with confidence intervals.
m_hr <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, hr = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
# 9. Continuous MD/SMD, single proportion/rate, correlation, and IRR.
cont <- data.frame(
study = paste0("C", 1:5),
m1 = c(12, 14, 13, 16, 15), s1 = c(3, 4, 3, 5, 4), n1 = rep(60, 5),
m0 = c(15, 15, 16, 18, 16), s0 = c(4, 4, 5, 5, 4), n0 = rep(60, 5)
)
m_md <- tabmeta(
cont, study, mean1 = m1, sd1 = s1, n1 = n1,
mean0 = m0, sd0 = s0, n0 = n0, md = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
m_smd <- tabmeta(
cont, study, mean1 = m1, sd1 = s1, n1 = n1,
mean0 = m0, sd0 = s0, n0 = n0, smd = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
one <- data.frame(
study = paste0("P", 1:5), events = c(8, 12, 15, 10, 14),
total = c(100, 110, 120, 90, 105), person_time = c(80, 90, 95, 75, 88),
correlation = c(.20, .28, .15, .31, .24)
)
m_prop <- tabmeta(
one, study, event = events, n = total, prop = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
m_rate <- tabmeta(
one, study, event = events, time = person_time, rate = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
m_cor <- tabmeta(
one, study, effect = correlation, n = total, cor = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
m_irr <- tabmeta(
dat[1:5, ], study,
event1 = event_treat, time1 = n_treat,
event0 = event_control, time0 = n_control, irr = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
# 10. Cumulative meta-analysis ordered by publication year.
m_cum <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
cumulative = year,
profile = "custom", plot = FALSE, show = FALSE
)
m_cum$tables$Cumulative
# 11. Draw or save individual publication figures.
if (interactive()) {
# plot=TRUE keeps all figures in the HTML Viewer and also draws the
# selected figures in the R/RStudio Plot pane and Plot history.
m_publication <- tabmeta(
data = dat, study = study,
a = event_treat, b = non_event_treat,
c = event_control, d = non_event_control,
interpretation = TRUE,
plot = TRUE,
plot_display = c("forest", "funnel", "trimfill"),
plot_args = list(
all = list(
font_family = "Arial", background = "white",
title_color = "#17365D", title_size = 1.1,
text_size = 0.86, axis_size = 0.92
),
forest = list(
subtitle = "Random-effects model with 95% confidence intervals",
caption = "Square size reflects study weight; diamond is pooled effect.",
margins = c(5.5, 4.2, 5.0, 2.0),
point_color = "#1F4E79", ci_color = "#5B9BD5",
summary_color = "#C00000", summary_border = "#7F0000",
point_shape = 15, row_shade = "zebra",
shade_color = "#F5F7FA", show_weights = TRUE,
weight_title = "Weight", estimate_title = "OR (95% CI)",
show_prediction = FALSE,
ref_color = "#666666", ref_type = 2,
xlim = c(0.2, 2.0), ticks = c(0.25, 0.5, 1, 1.5, 2)
),
funnel = list(
point_shape = 21, point_color = "#1F4E79",
point_bg = "#D9EAF7", point_size = 1.1,
contour_levels = c(90, 95, 99),
contour_colors = c("#FFF2CC", "#FCE4D6", "#E2F0D9"),
funnel_label = "out", funnel_legend = "topright"
)
)
)
m_publication$tables$Interpretation
# Any figure can be redrawn or saved independently.
plot(m_publication, type = "forest")
plot(m_publication, type = "funnel", contour = TRUE)
plot(m_publication, type = "trimfill", contour = TRUE)
plot(m_reg, type = "bubble", moderator = "year")
plot(
m_publication, type = "forest", file = "forest_publication.tiff",
width = 2400, height = 1800, res = 300,
font_family = "Arial", point_color = "#1F4E79",
ci_color = "#5B9BD5", summary_color = "#C00000",
show_weights = TRUE, show_prediction = FALSE,
xlim = c(0.5, 1.5), ticks = c(0.5, 0.75, 1, 1.25, 1.5)
)
# Values outside xlim remain exact in the Estimate (95% CI) column;
# the graphical confidence interval is clipped with an arrow.
m_publication$overall[c("pi_lower", "pi_upper")]
tabmeta(
data = dat, study = study,
a = event_treat, b = non_event_treat,
c = event_control, d = non_event_control,
export = c("docx", "xlsx"), file = "meta_report",
show = FALSE
)
}
}
R4VN documentation built on Sept. 30, 2026, 5:13 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
| tabmeta | R Documentation |
Publication-Ready Meta-Analysis in One Command
Description
Fits fixed- and/or random-effects meta-analysis and returns a complete,
publication-ready report. tabmeta() accepts a binary 2-by-2 table
(a, b, c, d), event/total data, continuous summaries, rates,
correlations, or study-level effect estimates with standard errors or
confidence intervals. The default profile="auto" adds prediction,
few-study inference, subgroup/moderator results when requested,
small-study-effect diagnostics, sensitivity analyses, and figures.
Usage
tabmeta(
data = NULL,
study,
effect = NULL,
se = NULL,
lower = NULL,
upper = NULL,
a = NULL,
b = NULL,
c = NULL,
d = NULL,
event1 = NULL,
n1 = NULL,
event0 = NULL,
n0 = NULL,
mean1 = NULL,
sd1 = NULL,
mean0 = NULL,
sd0 = NULL,
event = NULL,
n = NULL,
time = NULL,
time1 = NULL,
time0 = NULL,
or = FALSE,
rr = FALSE,
rd = FALSE,
hr = FALSE,
irr = FALSE,
md = FALSE,
smd = FALSE,
prop = FALSE,
rate = FALSE,
cor = FALSE,
fixed = FALSE,
random = TRUE,
method = "REML",
hk = NULL,
small = c("auto", "adhoc", "knha", "t", "z"),
prediction = NULL,
by = NULL,
subgroup = NULL,
reg = NULL,
moderator = NULL,
bias = NULL,
bias_methods = "auto",
leaveout = NULL,
influence = NULL,
cumulative = NULL,
transform = NULL,
cc = 0.5,
zero = c("keep", "exclude"),
ci = 0.95,
digit = 2,
p_digit = 3,
profile = c("auto", "brief", "full", "custom"),
full = NULL,
plot = NULL,
plot_display = "forest",
plot_args = list(),
report = FALSE,
interpretation = FALSE,
title = NULL,
show = TRUE,
export = NULL,
file = NULL,
open = FALSE,
strict = FALSE
)
Arguments
data |
Optional data frame. When omitted, active R4VN data are used. |
study |
Study label variable. |
effect |
Generic study-level effect estimate on the natural scale. |
se |
Standard error on the analysis scale. For ratio measures this is the standard error of the log effect. |
lower, upper |
Lower and upper confidence limits for |
a, b, c, d |
Binary 2-by-2 cells: events and non-events in group 1
( |
event1, n1, event0, n0 |
Events and total sample sizes in groups 1 and 0. |
mean1, sd1, mean0, sd0 |
Group means and standard deviations. |
event, n, time |
Single-group events, sample size, and person-time. |
time1, time0 |
Person-time in groups 1 and 0 for incidence-rate ratios. |
or, rr, rd, hr, irr, md, smd, prop, rate, cor |
Logical effect selectors. With raw binary input, OR is inferred if none is selected; with continuous summaries, MD is inferred. Set a selector to request another measure. |
fixed |
Fit a fixed-effect model in addition to, or instead of, the random-effects model. |
random |
Fit a random-effects model; default |
method |
Random-effects tau-squared estimator; default |
hk |
Backward-compatible logical shortcut. |
small |
Random-effects inference: |
prediction |
Add a prediction interval. |
by, subgroup |
One or more subgroup variables. Use one unquoted variable,
a character vector, or |
reg, moderator |
One or more meta-regression moderators created with
|
bias |
Assess small-study effects/publication bias. |
bias_methods |
One or more of |
leaveout |
Perform leave-one-out sensitivity analysis. |
influence |
Perform influence diagnostics. |
cumulative |
Optional variable defining the ordering for cumulative meta-analysis, usually publication year. |
transform |
Single-proportion transformation: |
cc |
Continuity correction for zero events; default |
zero |
Handling of double-zero binary studies: |
ci |
Confidence level as a proportion. |
digit |
Number of decimals for effect estimates and confidence limits,
including subgroup and meta-regression estimates. Default |
p_digit |
Number of decimals for p-values. |
profile |
Analysis profile. |
full |
Backward-compatible profile shortcut: |
plot |
Create publication-ready plots. |
plot_display |
Plot type(s) also drawn in the interactive R/RStudio Plot
pane when |
plot_args |
Named list of plot options. Supply common options directly,
or nested lists such as |
report |
Add manuscript-style results text. |
interpretation |
Add a detailed, sectioned interpretation covering the
pooled effect, heterogeneity, prediction interval, subgroup differences,
moderators, small-study effects, few-study inference, influence,
leave-one-out robustness, and cumulative evidence whenever available.
Default is |
title |
Report title. |
show |
Open the complete HTML report; default |
export |
Optional direct export format(s): |
file |
Export file or base path. The extension may determine the format. |
open |
Open the exported file. |
strict |
Stop when an optional diagnostic or figure cannot be created.
The default |
Value
Invisibly returns an object of class r4vn_meta and r4vn_tab.
Stable publication components are available in $estimates, $tests,
$diagnostics, $models, $tables, and $metadata. The table
$tables$Statistical_tests gives the pooled-effect and Cochran's Q tests.
Its significance and conclusion columns are added only when
interpretation=TRUE. $tables$Subgroup is the transposed publication
table for the first subgroup variable, $subgroup_tables contains every
transposed subgroup table, and $subgroup_long retains tidy long output.
See Also
tabexport, vars
Other R4VN tables:
tab(),
tabexport(),
tabforest(),
tablong(),
tabmulti(),
tabscale(),
tabscore(),
tabsurvey(),
vars()
Examples
if (requireNamespace("metafor", quietly = TRUE)) {
dat <- read.csv(
system.file("extdata", "meta_example.csv", package = "R4VN")
)
# 1. Binary outcome from a, b, c, d. OR is inferred automatically.
dat$non_event_treat <- dat$n_treat - dat$event_treat
dat$non_event_control <- dat$n_control - dat$event_control
m_abcd <- tabmeta(
data = dat, study = study,
a = event_treat, b = non_event_treat,
c = event_control, d = non_event_control,
profile = "custom", plot = FALSE, show = FALSE
)
m_abcd$estimates$overall
m_abcd$tables$Binary_2x2
m_abcd$tables$Statistical_tests
m_abcd$tests$overall
m_abcd$tests$heterogeneity
plot(
m_abcd, type = "forest", show_abcd = TRUE,
abcd_titles = c("Events T", "No event T", "Events C", "No event C")
)
# 2. Equivalent event/total syntax; request RR or RD with rr/rd=TRUE.
m_or <- tabmeta(
data = dat, study = study,
event1 = event_treat, n1 = n_treat,
event0 = event_control, n0 = n_control,
or = TRUE, profile = "custom", plot = FALSE, show = FALSE
)
# 3. Complete automatic analysis: report, prediction, diagnostics, plots.
m_all <- tabmeta(
data = dat, study = study,
a = event_treat, b = non_event_treat,
c = event_control, d = non_event_control,
show = FALSE
)
# 4. Subgroup analysis: one or several subgroup variables.
m_sub <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
subgroup = region,
profile = "custom", plot = FALSE, show = FALSE
)
m_sub$tables$Subgroup
m_sub$subgroup_long
m_sub$tables$Subgroup_test
# Each subgroup row includes its effect test and heterogeneity test.
names(m_sub$tables$Subgroup)
# If the subgroup variable carries value labels, R4VN prints those labels
# instead of numeric codes. With plot=TRUE, both Viewer and Plot history
# contain the overall plot and one clearly titled plot per subgroup level.
dat$risk_group <- rep(c(1, 2), length.out = nrow(dat))
attr(dat$risk_group, "label") <- "Baseline risk"
attr(dat$risk_group, "labels") <- c("Lower risk" = 1, "Higher risk" = 2)
m_labelled <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
subgroup = risk_group, profile = "custom",
plot = TRUE, plot_display = "forest", show = FALSE
)
m_labelled$plot_titles
# Multiple subgroup analyses can be requested together.
dat$period <- ifelse(dat$year < median(dat$year), "Earlier", "Later")
m_sub2 <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
subgroup = vars(region, period),
profile = "custom", plot = FALSE, show = FALSE
)
# 5. Meta-regression with numeric and categorical moderators.
attr(dat$year, "label") <- "Publication year"
attr(dat$region, "label") <- "Geographic region"
m_reg <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
moderator = vars(c.year, region),
profile = "custom", plot = FALSE, show = FALSE
)
m_reg$tables$Meta_regression
m_reg$tables$Moderator_univariable
# Intercept is written in full; moderator labels are used when available.
# Set digit=4, for example, when four decimal places are required.
# 6. Publication-bias and small-study-effect sensitivity analyses.
m_bias <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
bias = TRUE,
bias_methods = c("egger", "begg", "trimfill", "failsafe"),
profile = "custom", plot = FALSE, show = FALSE
)
m_bias$tables$Publication_bias
m_bias$tables$Publication_bias_adjusted
# 7. Few-study inference. auto uses modified Hartung-Knapp at <=10 studies.
m_few <- tabmeta(
data = dat[1:8, ], study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
small = "auto", profile = "custom", plot = FALSE, show = FALSE
)
m_few$tables$Small_sample_inference
# 8. Generic hazard ratios with confidence intervals.
m_hr <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, hr = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
# 9. Continuous MD/SMD, single proportion/rate, correlation, and IRR.
cont <- data.frame(
study = paste0("C", 1:5),
m1 = c(12, 14, 13, 16, 15), s1 = c(3, 4, 3, 5, 4), n1 = rep(60, 5),
m0 = c(15, 15, 16, 18, 16), s0 = c(4, 4, 5, 5, 4), n0 = rep(60, 5)
)
m_md <- tabmeta(
cont, study, mean1 = m1, sd1 = s1, n1 = n1,
mean0 = m0, sd0 = s0, n0 = n0, md = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
m_smd <- tabmeta(
cont, study, mean1 = m1, sd1 = s1, n1 = n1,
mean0 = m0, sd0 = s0, n0 = n0, smd = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
one <- data.frame(
study = paste0("P", 1:5), events = c(8, 12, 15, 10, 14),
total = c(100, 110, 120, 90, 105), person_time = c(80, 90, 95, 75, 88),
correlation = c(.20, .28, .15, .31, .24)
)
m_prop <- tabmeta(
one, study, event = events, n = total, prop = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
m_rate <- tabmeta(
one, study, event = events, time = person_time, rate = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
m_cor <- tabmeta(
one, study, effect = correlation, n = total, cor = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
m_irr <- tabmeta(
dat[1:5, ], study,
event1 = event_treat, time1 = n_treat,
event0 = event_control, time0 = n_control, irr = TRUE,
profile = "custom", plot = FALSE, show = FALSE
)
# 10. Cumulative meta-analysis ordered by publication year.
m_cum <- tabmeta(
data = dat, study = study,
effect = OR, lower = LCI, upper = UCI, or = TRUE,
cumulative = year,
profile = "custom", plot = FALSE, show = FALSE
)
m_cum$tables$Cumulative
# 11. Draw or save individual publication figures.
if (interactive()) {
# plot=TRUE keeps all figures in the HTML Viewer and also draws the
# selected figures in the R/RStudio Plot pane and Plot history.
m_publication <- tabmeta(
data = dat, study = study,
a = event_treat, b = non_event_treat,
c = event_control, d = non_event_control,
interpretation = TRUE,
plot = TRUE,
plot_display = c("forest", "funnel", "trimfill"),
plot_args = list(
all = list(
font_family = "Arial", background = "white",
title_color = "#17365D", title_size = 1.1,
text_size = 0.86, axis_size = 0.92
),
forest = list(
subtitle = "Random-effects model with 95% confidence intervals",
caption = "Square size reflects study weight; diamond is pooled effect.",
margins = c(5.5, 4.2, 5.0, 2.0),
point_color = "#1F4E79", ci_color = "#5B9BD5",
summary_color = "#C00000", summary_border = "#7F0000",
point_shape = 15, row_shade = "zebra",
shade_color = "#F5F7FA", show_weights = TRUE,
weight_title = "Weight", estimate_title = "OR (95% CI)",
show_prediction = FALSE,
ref_color = "#666666", ref_type = 2,
xlim = c(0.2, 2.0), ticks = c(0.25, 0.5, 1, 1.5, 2)
),
funnel = list(
point_shape = 21, point_color = "#1F4E79",
point_bg = "#D9EAF7", point_size = 1.1,
contour_levels = c(90, 95, 99),
contour_colors = c("#FFF2CC", "#FCE4D6", "#E2F0D9"),
funnel_label = "out", funnel_legend = "topright"
)
)
)
m_publication$tables$Interpretation
# Any figure can be redrawn or saved independently.
plot(m_publication, type = "forest")
plot(m_publication, type = "funnel", contour = TRUE)
plot(m_publication, type = "trimfill", contour = TRUE)
plot(m_reg, type = "bubble", moderator = "year")
plot(
m_publication, type = "forest", file = "forest_publication.tiff",
width = 2400, height = 1800, res = 300,
font_family = "Arial", point_color = "#1F4E79",
ci_color = "#5B9BD5", summary_color = "#C00000",
show_weights = TRUE, show_prediction = FALSE,
xlim = c(0.5, 1.5), ticks = c(0.5, 0.75, 1, 1.25, 1.5)
)
# Values outside xlim remain exact in the Estimate (95% CI) column;
# the graphical confidence interval is clipped with an arrow.
m_publication$overall[c("pi_lower", "pi_upper")]
tabmeta(
data = dat, study = study,
a = event_treat, b = non_event_treat,
c = event_control, d = non_event_control,
export = c("docx", "xlsx"), file = "meta_report",
show = FALSE
)
}
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.