Forest plots for dichotomous outcomes and ‘O–E and
Variance’ outcomes illustrate, by default:
the
raw data (corresponding to the 2´2 tables) for each study;
point
estimates and confidence intervals for the chosen effect measure, both
as blocks and lines and as text;
a meta-analysis
for each subgroup using the chosen effect measure and chosen method (fixed
or random effects), both as a diamond and as text
the
total numbers of participants and total numbers with events in the experimental
intervention and control intervention groups;
heterogeneity statistics (among-study
variance (tau-squared, or Tau2, or
τ2) for random-effects meta-analyses,
the chi-squared test, the I2 statistic
and a test for differences across subgroups if they are present and appropriate);
a test
for overall effect (overall average effect for random-effects meta-analyses);
and
percent
weights given to each study.
Note that 3–7
are not displayed unless data are pooled. Furthermore, the test for differences
across subgroups is not displayed for Mantel-Haenszel analyses. For ‘O–E
and Variance’ outcomes it is also possible to enable display of the O–E
and V statistics.
Forest plots for continuous outcomes illustrate, by
default:
the
raw data (means, standard deviations and sample sizes) for each arm in
each study;
point
estimates and confidence intervals for the chosen effect measure, both
as blocks and lines and as text;
a meta-analysis
for each subgroup using the chosen effect measure and chosen method (fixed
or random effects), both as a diamond and as text;
the
total numbers of participants in the experimental and control groups;
heterogeneity
statistics (among-study variance (tau-squared) for random-effects meta-analyses,
the chi-squared test, the I2 statistic and a test for differences across subgroups if they
are present);
a test
for overall effect (overall average effect for random-effects meta-analyses);
and
percent
weights given to each study.
Note that 3–7
are not displayed unless the data are pooled.
Forest plots for the generic inverse variance method
illustrate, by default:
the
summary data for each study, as entered by the author (for ratio measures
these will be on the natural log (‘ln’) scale);
point
estimates and confidence intervals, both as blocks and lines and as text
(for ratio measures these will be on the natural scale rather than the
log scale);
a meta-analysis
for each subgroup using the chosen method (fixed or random effects), both
as a diamond and as text;
heterogeneity
statistics (among-study variance (tau-squared) for random-effects meta-analyses,
the chi-squared test, the I2 statistic, and a test for differences
across subgroups if they are present);
a test
for overall effect (overall average effect for random-effects meta-analyses);
and
percent
weights given to each study.
Note that 3–6
are not shown unless data are pooled. It is possible additionally to enter
sample sizes for experimental and control groups. These should be entered
as appropriate for the design of the study. The sample sizes are not involved
in the analysis, but if entered are displayed as:
numbers
of participants in the experimental and control group for each study;
the
total numbers of participants in the experimental and control groups.
|