Skip to contents

Display treatment effects against the fitted reference by default, with an explicit option for all pairwise comparisons.

Usage

# S3 method for class 'nma_coherent'
print(
  x,
  ...,
  comparisons = c("reference", "all"),
  digits = getOption("digits"),
  detailed = FALSE
)

# S3 method for class 'nma_coherent'
summary(
  object,
  ...,
  comparisons = c("reference", "all"),
  digits = getOption("digits"),
  detailed = FALSE,
  inference = NULL,
  test = NULL,
  robust_type = NULL,
  cluster = NULL,
  hksj_adjust = NULL,
  level = NULL
)

Arguments

x, object

A fitted nma_coherent network model.

...

Reserved additional print or summary arguments.

comparisons

reference (default) displays each treatment minus the fitted reference, within every mean stratum. all displays all pairwise contrasts in the stored treatment_2-minus-treatment_1 direction.

digits

Number of decimal places to display, an integer from 1 to 22. Defaults to getOption("digits"). A summary remembers the requested precision; an explicit print(..., digits=) overrides it. This controls formatting only, not stored estimates or calculations. Counts and df remain integers and outcome/time identifiers remain unrounded. Trailing zeros are retained. Raw and adjusted p-values below 0.0001 display as <0.0001; other p-values use at least four decimal places (max(4, digits)). Full precision remains available in $fit_coefficients and $effects.

detailed

Logical; show the complete coefficient table as well as the selected treatment contrasts. Defaults to FALSE: basic treatment coefficients are omitted, while moderator coefficients remain visible. A summary remembers this choice; explicit print(..., detailed=) overrides it.

inference

Optional summary-only model or robust inference framework.

test

Optional summary-only coefficient/total-contrast test: z or Satterthwaite; robust inference also accepts naive-t and naive-tp.

robust_type

Optional summary-only sandwich covariance, e.g. CR2.

cluster

Optional fitted contrast-data cluster column, default study_id.

hksj_adjust

Unsupported for networks; supplied values raise an error.

level

Optional summary-only coefficient/total-contrast confidence level.

Value

print() invisibly returns its input. summary() returns the fitted object with display_comparisons, display_digits and display_detailed attributes for subsequent printing; its complete model results and classes are retained.

Details

Printed treatment comparisons have a contrast label of treatment_2 minus treatment_1 (for example, Metformin - Placebo). Stored treatment identifiers and estimates retain their original structure. Reference display sets treatment_1 to the fitted reference and treatment_2 to the other treatment. Estimates are negated and interval endpoints reversed when the stored contrast has the opposite direction. Standard errors and p-values are unchanged. Coefficients and variance summaries are displayed as fitted. Moderator-zero evaluation and stratum definitions are unchanged. Basic treatment coefficients reproduce the reference contrasts at moderator zero. The default display therefore shows treatment results once, as contrasts. For moderator models, additional interaction coefficients are displayed in a separate table. detailed=TRUE restores all basic and moderator coefficients. Treatment contrasts include statistic = estimate / SE; its reference distribution (normal or t with the fitted df) is stated in the model settings.

Display selection does not refit the model or restrict $effects, league tables, confidence-interval methods, or incoherence diagnostic hypotheses. Multiplicity-adjusted p-values retain the original fitted family, including comparisons omitted from the display. This is not an adjustment over only reference comparisons.

summary(object, comparisons="all") records the display choice on the returned copy. An explicit print(..., comparisons=) overrides that choice. The input fitted object is unchanged. Network counts use analyzed studies and their full treatment sets. Pairwise comparisons count each unordered pair once per study, including multi-arm comparisons implied by an independent contrast basis. Unique treatment pairs count those pairs once across studies. Repeated outcomes/times do not create additional studies or duplicate study pairs. Modeled contrast rows count the actual analyzed rows; they may be correlated. Designs are unique study treatment sets. Counts are independent of the reference and display selection.

Residual variation uses network_q_decomposition(): sampling-covariance GLS overall, within-design and between-design Q with residual/rank-increase df. Overall Q reflects both within-design heterogeneity and between-design incoherence. Q-based I-squared is 100 * max(0, 1 - df / Q) for the overall and within-design components when df is positive; Q=0 gives zero. With zero df it is unavailable, not zero. Between-design I-squared is not reported. These I-squared values are not additive, do not partition multivariate or multilevel variance components, and do not use fitted random-effect covariance. Shared moderators remain in the mean model, so these are residual diagnostics. The random-effects Higgins design test is a separate assessment. Numerical failure of the sampling-only calculation is labelled unavailable and does not prevent printing the fitted model.

Summary-only inference='model' or inference='robust' can select model normal, expected-information Satterthwaite, or cluster-robust inference. Model Satterthwaite requires converged interior estimated heterogeneity. HKSJ is unsupported. These overrides affect basic coefficients and total treatment contrasts only; all other calculations retain the fitted settings.

See also

Examples

d <- data.frame(s=1:9,a=rep(c("A","A","B"),3),
  b=rep(c("B","C","C"),3),g=rep(c(.2,.6,.2),3),v=.04)
fit <- network_meta_coherent(d,"s","a","b","g","v",reference="B")
print(fit)
#> Results generated with MARS: v0.6.0.8
#> Printed on: Friday, November 20, 2026
#> 
#> Joint coherent network meta-analysis
#> ====================================
#> 
#> Network overview
#>   Studies:                         9
#>   Pairwise comparisons:            9
#>   Treatments:                      3
#>   Designs:                         3
#>   Unique study treatment pairs:    3
#>   Modeled contrast rows:           9
#>   Pairwise comparisons count each treatment pair once per study,
#>   including implied multi-arm pairs; unique pairs are counted once overall.
#> 
#> Model settings
#>   Interface:                       coherent
#>   Reference treatment:             B
#>   Effect direction:                Treatment minus B
#>   Estimation:                      REML
#>   Sampling covariance:             independent
#>   Heterogeneity structure:         none
#>   Inference:                       model
#>   Test reference distribution:     Normal (z)
#>   Convergence code:                0
#>   Covariance boundary:             TRUE
#>   Treatment p-value adjustment:    none
#>   Adjustment family size:          3
#> 
#> Residual variation (sampling covariance only)
#>        Component         Q df   p-value    I2 (%)
#>          Overall 1.0000000  7 0.9948285 0.0000000
#>   Within designs 0.0000000  6 1.0000000 0.0000000
#>  Between designs 1.0000000  1 0.3173105        NA
#>   Overall Q/I2 include within-design heterogeneity and between-design incoherence.
#>   I2 is Q-based (%); overall and within-design I2 are not additive.
#>   These are sampling-only diagnostics, separate from the random-effects Higgins test.
#> 
#> Fitted variance components
#>  component      tau2
#>      study 0.0000000
#> 
#> Treatment contrasts (treatment minus B; reference comparisons):
#> Raw p-values; confidence intervals: pointwise.
#>  contrast      total  total_se  statistic total_ci_lower total_ci_upper
#>     A - B -0.2666667 0.0942809 -2.8284271     -0.4514538     -0.0818795
#>     C - B  0.2666667 0.0942809  2.8284271      0.0818795      0.4514538
#>  total_p_value total_p_adjusted
#>      0.0046777        0.0046777
#>      0.0046777        0.0046777
summary(fit, comparisons="all")
#> Results generated with MARS: v0.6.0.8
#> Printed on: Friday, November 20, 2026
#> 
#> Joint coherent network meta-analysis
#> ====================================
#> 
#> Network overview
#>   Studies:                         9
#>   Pairwise comparisons:            9
#>   Treatments:                      3
#>   Designs:                         3
#>   Unique study treatment pairs:    3
#>   Modeled contrast rows:           9
#>   Pairwise comparisons count each treatment pair once per study,
#>   including implied multi-arm pairs; unique pairs are counted once overall.
#> 
#> Model settings
#>   Interface:                       coherent
#>   Reference treatment:             B
#>   Effect direction:                treatment_2 minus treatment_1
#>   Estimation:                      REML
#>   Sampling covariance:             independent
#>   Heterogeneity structure:         none
#>   Inference:                       model
#>   Test reference distribution:     Normal (z)
#>   Convergence code:                0
#>   Covariance boundary:             TRUE
#>   Treatment p-value adjustment:    none
#>   Adjustment family size:          3
#> 
#> Residual variation (sampling covariance only)
#>        Component         Q df   p-value    I2 (%)
#>          Overall 1.0000000  7 0.9948285 0.0000000
#>   Within designs 0.0000000  6 1.0000000 0.0000000
#>  Between designs 1.0000000  1 0.3173105        NA
#>   Overall Q/I2 include within-design heterogeneity and between-design incoherence.
#>   I2 is Q-based (%); overall and within-design I2 are not additive.
#>   These are sampling-only diagnostics, separate from the random-effects Higgins test.
#> 
#> Fitted variance components
#>  component      tau2
#>      study 0.0000000
#> 
#> Treatment contrasts (all pairs; treatment_2 minus treatment_1):
#> Raw p-values; confidence intervals: pointwise.
#>  contrast     total  total_se statistic total_ci_lower total_ci_upper
#>     B - A 0.2666667 0.0942809 2.8284271      0.0818795      0.4514538
#>     C - A 0.5333333 0.0942809 5.6568542      0.3485462      0.7181205
#>     C - B 0.2666667 0.0942809 2.8284271      0.0818795      0.4514538
#>  total_p_value total_p_adjusted
#>      0.0046777        0.0046777
#>        <0.0001          <0.0001
#>      0.0046777        0.0046777
print(fit, comparisons="all")
#> Results generated with MARS: v0.6.0.8
#> Printed on: Friday, November 20, 2026
#> 
#> Joint coherent network meta-analysis
#> ====================================
#> 
#> Network overview
#>   Studies:                         9
#>   Pairwise comparisons:            9
#>   Treatments:                      3
#>   Designs:                         3
#>   Unique study treatment pairs:    3
#>   Modeled contrast rows:           9
#>   Pairwise comparisons count each treatment pair once per study,
#>   including implied multi-arm pairs; unique pairs are counted once overall.
#> 
#> Model settings
#>   Interface:                       coherent
#>   Reference treatment:             B
#>   Effect direction:                treatment_2 minus treatment_1
#>   Estimation:                      REML
#>   Sampling covariance:             independent
#>   Heterogeneity structure:         none
#>   Inference:                       model
#>   Test reference distribution:     Normal (z)
#>   Convergence code:                0
#>   Covariance boundary:             TRUE
#>   Treatment p-value adjustment:    none
#>   Adjustment family size:          3
#> 
#> Residual variation (sampling covariance only)
#>        Component         Q df   p-value    I2 (%)
#>          Overall 1.0000000  7 0.9948285 0.0000000
#>   Within designs 0.0000000  6 1.0000000 0.0000000
#>  Between designs 1.0000000  1 0.3173105        NA
#>   Overall Q/I2 include within-design heterogeneity and between-design incoherence.
#>   I2 is Q-based (%); overall and within-design I2 are not additive.
#>   These are sampling-only diagnostics, separate from the random-effects Higgins test.
#> 
#> Fitted variance components
#>  component      tau2
#>      study 0.0000000
#> 
#> Treatment contrasts (all pairs; treatment_2 minus treatment_1):
#> Raw p-values; confidence intervals: pointwise.
#>  contrast     total  total_se statistic total_ci_lower total_ci_upper
#>     B - A 0.2666667 0.0942809 2.8284271      0.0818795      0.4514538
#>     C - A 0.5333333 0.0942809 5.6568542      0.3485462      0.7181205
#>     C - B 0.2666667 0.0942809 2.8284271      0.0818795      0.4514538
#>  total_p_value total_p_adjusted
#>      0.0046777        0.0046777
#>        <0.0001          <0.0001
#>      0.0046777        0.0046777
summary(fit, digits=3)
#> Results generated with MARS: v0.6.0.8
#> Printed on: Friday, November 20, 2026
#> 
#> Joint coherent network meta-analysis
#> ====================================
#> 
#> Network overview
#>   Studies:                         9
#>   Pairwise comparisons:            9
#>   Treatments:                      3
#>   Designs:                         3
#>   Unique study treatment pairs:    3
#>   Modeled contrast rows:           9
#>   Pairwise comparisons count each treatment pair once per study,
#>   including implied multi-arm pairs; unique pairs are counted once overall.
#> 
#> Model settings
#>   Interface:                       coherent
#>   Reference treatment:             B
#>   Effect direction:                Treatment minus B
#>   Estimation:                      REML
#>   Sampling covariance:             independent
#>   Heterogeneity structure:         none
#>   Inference:                       model
#>   Test reference distribution:     Normal (z)
#>   Convergence code:                0
#>   Covariance boundary:             TRUE
#>   Treatment p-value adjustment:    none
#>   Adjustment family size:          3
#> 
#> Residual variation (sampling covariance only)
#>        Component     Q df p-value I2 (%)
#>          Overall 1.000  7  0.9948  0.000
#>   Within designs 0.000  6  1.0000  0.000
#>  Between designs 1.000  1  0.3173     NA
#>   Overall Q/I2 include within-design heterogeneity and between-design incoherence.
#>   I2 is Q-based (%); overall and within-design I2 are not additive.
#>   These are sampling-only diagnostics, separate from the random-effects Higgins test.
#> 
#> Fitted variance components
#>  component  tau2
#>      study 0.000
#> 
#> Treatment contrasts (treatment minus B; reference comparisons):
#> Raw p-values; confidence intervals: pointwise.
#>  contrast  total total_se statistic total_ci_lower total_ci_upper total_p_value
#>     A - B -0.267    0.094    -2.828         -0.451         -0.082        0.0047
#>     C - B  0.267    0.094     2.828          0.082          0.451        0.0047
#>  total_p_adjusted
#>            0.0047
#>            0.0047
print(summary(fit, digits=3), digits=5)
#> Results generated with MARS: v0.6.0.8
#> Printed on: Friday, November 20, 2026
#> 
#> Joint coherent network meta-analysis
#> ====================================
#> 
#> Network overview
#>   Studies:                         9
#>   Pairwise comparisons:            9
#>   Treatments:                      3
#>   Designs:                         3
#>   Unique study treatment pairs:    3
#>   Modeled contrast rows:           9
#>   Pairwise comparisons count each treatment pair once per study,
#>   including implied multi-arm pairs; unique pairs are counted once overall.
#> 
#> Model settings
#>   Interface:                       coherent
#>   Reference treatment:             B
#>   Effect direction:                Treatment minus B
#>   Estimation:                      REML
#>   Sampling covariance:             independent
#>   Heterogeneity structure:         none
#>   Inference:                       model
#>   Test reference distribution:     Normal (z)
#>   Convergence code:                0
#>   Covariance boundary:             TRUE
#>   Treatment p-value adjustment:    none
#>   Adjustment family size:          3
#> 
#> Residual variation (sampling covariance only)
#>        Component       Q df p-value  I2 (%)
#>          Overall 1.00000  7 0.99483 0.00000
#>   Within designs 0.00000  6 1.00000 0.00000
#>  Between designs 1.00000  1 0.31731      NA
#>   Overall Q/I2 include within-design heterogeneity and between-design incoherence.
#>   I2 is Q-based (%); overall and within-design I2 are not additive.
#>   These are sampling-only diagnostics, separate from the random-effects Higgins test.
#> 
#> Fitted variance components
#>  component    tau2
#>      study 0.00000
#> 
#> Treatment contrasts (treatment minus B; reference comparisons):
#> Raw p-values; confidence intervals: pointwise.
#>  contrast    total total_se statistic total_ci_lower total_ci_upper
#>     A - B -0.26667  0.09428  -2.82843       -0.45145       -0.08188
#>     C - B  0.26667  0.09428   2.82843        0.08188        0.45145
#>  total_p_value total_p_adjusted
#>        0.00468          0.00468
#>        0.00468          0.00468
summary(fit, detailed=TRUE)
#> Results generated with MARS: v0.6.0.8
#> Printed on: Friday, November 20, 2026
#> 
#> Joint coherent network meta-analysis
#> ====================================
#> 
#> Network overview
#>   Studies:                         9
#>   Pairwise comparisons:            9
#>   Treatments:                      3
#>   Designs:                         3
#>   Unique study treatment pairs:    3
#>   Modeled contrast rows:           9
#>   Pairwise comparisons count each treatment pair once per study,
#>   including implied multi-arm pairs; unique pairs are counted once overall.
#> 
#> Model settings
#>   Interface:                       coherent
#>   Reference treatment:             B
#>   Effect direction:                Treatment minus B
#>   Estimation:                      REML
#>   Sampling covariance:             independent
#>   Heterogeneity structure:         none
#>   Inference:                       model
#>   Test reference distribution:     Normal (z)
#>   Convergence code:                0
#>   Covariance boundary:             TRUE
#>   Treatment p-value adjustment:    none
#>   Adjustment family size:          3
#> 
#> Residual variation (sampling covariance only)
#>        Component         Q df   p-value    I2 (%)
#>          Overall 1.0000000  7 0.9948285 0.0000000
#>   Within designs 0.0000000  6 1.0000000 0.0000000
#>  Between designs 1.0000000  1 0.3173105        NA
#>   Overall Q/I2 include within-design heterogeneity and between-design incoherence.
#>   I2 is Q-based (%); overall and within-design I2 are not additive.
#>   These are sampling-only diagnostics, separate from the random-effects Higgins test.
#> 
#> Fitted variance components
#>  component      tau2
#>      study 0.0000000
#> 
#> Coefficient tests: raw p-values; confidence intervals: pointwise
#>  term   estimate std_error  df   se_source    z_value   p_value   ci_lower
#>   d_A -0.2666667 0.0942809 Inf model_based -2.8284271 0.0046777 -0.4514538
#>   d_C  0.2666667 0.0942809 Inf model_based  2.8284271 0.0046777  0.0818795
#>    ci_upper
#>  -0.0818795
#>   0.4514538
#> 
#> Treatment contrasts (treatment minus B; reference comparisons):
#> Raw p-values; confidence intervals: pointwise.
#>  contrast      total  total_se  statistic total_ci_lower total_ci_upper
#>     A - B -0.2666667 0.0942809 -2.8284271     -0.4514538     -0.0818795
#>     C - B  0.2666667 0.0942809  2.8284271      0.0818795      0.4514538
#>  total_p_value total_p_adjusted
#>      0.0046777        0.0046777
#>      0.0046777        0.0046777
print(summary(fit, detailed=TRUE), detailed=FALSE)
#> Results generated with MARS: v0.6.0.8
#> Printed on: Friday, November 20, 2026
#> 
#> Joint coherent network meta-analysis
#> ====================================
#> 
#> Network overview
#>   Studies:                         9
#>   Pairwise comparisons:            9
#>   Treatments:                      3
#>   Designs:                         3
#>   Unique study treatment pairs:    3
#>   Modeled contrast rows:           9
#>   Pairwise comparisons count each treatment pair once per study,
#>   including implied multi-arm pairs; unique pairs are counted once overall.
#> 
#> Model settings
#>   Interface:                       coherent
#>   Reference treatment:             B
#>   Effect direction:                Treatment minus B
#>   Estimation:                      REML
#>   Sampling covariance:             independent
#>   Heterogeneity structure:         none
#>   Inference:                       model
#>   Test reference distribution:     Normal (z)
#>   Convergence code:                0
#>   Covariance boundary:             TRUE
#>   Treatment p-value adjustment:    none
#>   Adjustment family size:          3
#> 
#> Residual variation (sampling covariance only)
#>        Component         Q df   p-value    I2 (%)
#>          Overall 1.0000000  7 0.9948285 0.0000000
#>   Within designs 0.0000000  6 1.0000000 0.0000000
#>  Between designs 1.0000000  1 0.3173105        NA
#>   Overall Q/I2 include within-design heterogeneity and between-design incoherence.
#>   I2 is Q-based (%); overall and within-design I2 are not additive.
#>   These are sampling-only diagnostics, separate from the random-effects Higgins test.
#> 
#> Fitted variance components
#>  component      tau2
#>      study 0.0000000
#> 
#> Treatment contrasts (treatment minus B; reference comparisons):
#> Raw p-values; confidence intervals: pointwise.
#>  contrast      total  total_se  statistic total_ci_lower total_ci_upper
#>     A - B -0.2666667 0.0942809 -2.8284271     -0.4514538     -0.0818795
#>     C - B  0.2666667 0.0942809  2.8284271      0.0818795      0.4514538
#>  total_p_value total_p_adjusted
#>      0.0046777        0.0046777
#>      0.0046777        0.0046777