thresh_forest: Producing threshold forest plots
In nmathresh: Thresholds and Invariant Intervals for Network Meta-Analysis

Description Usage Arguments Value Examples

This function produces threshold forest plots, overlaying the decision-invariant intervals on the data points and their confidence/credible intervals, as shown by Phillippo et al. (2018).

thresh_forest(
  thresh,
  y,
  CI.lo,
  CI.hi,
  label,
  orderby = NULL,
  data = NULL,
  CI.title = "95% Confidence Interval",
  label.title = "",
  y.title = "Mean",
  II.title = "Invariant Interval",
  xlab = "",
  xlim = NULL,
  sigfig = 3,
  digits = NULL,
  refline = NULL,
  clinsig = NULL,
  cutoff = NULL,
  II.colw = rgb(0.72, 0.8, 0.93),
  II.cols = rgb(0.93, 0.72, 0.8),
  II.lwd = 8,
  CI.lwd = 1,
  pointsize = 4,
  fontsize = 12,
  xbreaks = NULL,
  add.columns = NULL,
  add.columns.title = NULL,
  add.columns.after = -1,
  add.columns.hjust = 0.5,
  add.columns.uline = TRUE,
  calcdim = display,
  display = TRUE
)

`thresh`	A `thresh` object, as produced by `nma_thresh`.
`y`	Data points. Either a column of `data`, or a numeric vector.
`CI.lo`	Confidence/credible interval lower limits. Either a column of `data`, or a numeric vector.
`CI.hi`	Confidence/credible interval upper limits. Either a column of `data`, or a numeric vector.
`label`	Row labels (for each data point). Either a column of `data`, or a character vector.
`orderby`	Variable(s) to order the table rows by. Either a column or columns of `data`, or a vector. By default, the rows are not reordered. Further arguments and/or multiple ordering columns may be passed to the function `order` by instead providing a `list` containing the arguments to `order`.
`data`	A data frame containing the data points `y`, confidence/credible intervals (`CI.lo`, `CI.hi`), and row labels `labels`. If `data` is not provided, the above variables will be searched for in the calling environment.
`CI.title`	Title for CI column, default "95% Confidence Interval".
`label.title`	Character string giving the heading for the row labels column.
`y.title`	Character string giving the heading for the data points column, default "Mean".
`II.title`	Title for invariant interval column, default "Invariant Interval".
`xlab`	Character string giving the label for the x-axis.
`xlim`	Numeric vector (length 2) of lower and upper limits for the x-axis. If not set, tries to choose a sensible default.
`sigfig`	Number of significant digits to display in the table. Default 3.
`digits`	Number of decimal places to display in the table. Overrides `sigfig` if set.
`refline`	x intercept of vertical reference line, if desired.
`clinsig`	Set the clinical significance level. Rows are marked with a dagger if they have one or more thresholds less than this value. Not set by default.
`cutoff`	A single numeric value or numeric vector pair. Thresholds larger in magnitude than this value, or lying outside this interval, will be cut off and marked as NT (no threshold). Not set by default.
`II.colw`	Colour for "wide" invariant intervals.
`II.cols`	Colour for "short" invariant intervals.
`II.lwd`	Line width of invariant intervals. Default 8.
`CI.lwd`	Line width of confidence/credible intervals. Default 1.
`pointsize`	Point size for forest plot means. Default 4.
`fontsize`	Base font size. Default 12.
`xbreaks`	Position of tick marks on the x-axis as a numeric vector.
`add.columns`	Data frame (or matrix, vector) of additional columns to add to table.
`add.columns.title`	Optional titles for the additional columns, otherwise use names from provided data.
`add.columns.after`	Which column to add the new columns after? Default adds the columns to the far right.
`add.columns.hjust`	Vector of horizontal justifications for the new columns, from `0` (left) to `1` (right). Default centres every column.
`add.columns.uline`	Underline the header of the new columns? Default `TRUE`.
`calcdim`	Logical, calculate suggested output dimensions for saving to pdf? Calculates output size when `TRUE`; saves time when `FALSE`.
`display`	Logical, display the plot? Defaults to `TRUE`.

Displays the forest plot on the current plot device (if display = TRUE). Also returns invisibly the underlying gtable object, which can be further manipulated.

# Please see the vignette "Examples" for worked examples including use of
# this function, including more information on the brief code below.

vignette("Examples", package = "nmathresh")

### Contrast level thresholds for Thrombolytic treatments NMA
K <- 6   # Number of treatments

# Contrast design matrix is
X <- matrix(ncol = K-1, byrow = TRUE,
            c(1, 0, 0, 0, 0,
              0, 1, 0, 0, 0,
              0, 0, 1, 0, 0,
              0, 0, 0, 1, 0,
              0, -1, 1, 0, 0,
              0, -1, 0, 1, 0,
              0, -1, 0, 0, 1))

# Reconstruct hypothetical likelihood covariance matrix using NNLS
lik.cov <- recon_vcov(Thrombo.post.cov, prior.prec = .0001, X = X)

# Thresholds are then
thresh <- nma_thresh(mean.dk = Thrombo.post.summary$statistics[1:(K-1), "Mean"],
                     lhood = lik.cov,
                     post = Thrombo.post.cov,
                     nmatype = "fixed",
                     X = X,
                     opt.max = FALSE)

# Get treatment codes for the contrasts with data
d.a <- d.b <- vector(length = nrow(X))
for (i in 1:nrow(X)){
  d.a[i] <- ifelse(any(X[i, ] == -1), which(X[i, ] == -1), 0) + 1
  d.b[i] <- ifelse(any(X[i, ] == 1), which(X[i, ] == 1), 0) + 1
}

# Transform from d_ab style contrast references into d[i] style from the full set of contrasts
# for easy indexing in R
d.i <- d_ab2i(d.a, d.b, K = 6)

# Create plot data
plotdat <- data.frame(lab = paste0(d.b, " vs. ", d.a),
                      contr.mean = Thrombo.post.summary$statistics[d.i, "Mean"],
                      CI2.5 = Thrombo.post.summary$quantiles[d.i, "2.5%"],
                      CI97.5 = Thrombo.post.summary$quantiles[d.i, "97.5%"])

# Plot
thresh_forest(thresh, contr.mean, CI2.5, CI97.5, label = lab, data = plotdat,
              label.title = "Contrast", xlab = "Log Odds Ratio", CI.title = "95% Credible Interval",
              xlim = c(-.3, .3), refline = 0, digits = 2)