ltfu.fi: LTFU-aware fragility index calculator
In brb225/FragilityTools: Toolbox For Patient Count Fragility Measures

Description Usage Arguments Value Examples

This function implements a calculation of the LTFU-aware fragility indices described in LTFU fragility index. These fragility indices consider the potential impact of patients who are lost to follow up.

ltfu.fi(
  crosstab,
  ltfu,
  get.p,
  q = 0,
  alpha = 0.05,
  sampling.control = NULL,
  xlab = "Control group LTFU event rate",
  ylab = "Treatment group LTFU event rate",
  fig.size = 1.1,
  gradientn.scale = 0.99,
  effect.size.plt = FALSE,
  extreme = 1
)

`crosstab`	a 2x2 real non-negative integer matrix (contingency table) with control/treatment on the rows and event/nonevent on the columns.
`ltfu`	a length 2 integer vector counting the number of lost patients in both groups. The first entry corresponds to the first row of mat.
`get.p`	a function which inputs a 2x2 matrix and returns a p value
`q`	a numeric vector representing the sufficiently likely threshold. A LTFU-aware fragility index will be calculated for each.
`alpha`	a number for the significance threshold. The default is 0.05
`sampling.control`	a list providing the ndraws and mult arguments for sampling. It is NULL by default. When NULL, p_o = p_l is taken by default and the distribution of X_l \| X_o is available exactly. ndraws is the number of posterior draws, and mult is the multiplier s (see the LTFU-aware paper).
`xlab`	a string for the horizontal axis label of the output plot
`ylab`	a string for the vertical axis label of the output plot
`fig.size`	A number, by default 1.1. This gives the size of the tiles in the plot. It may need to be manually tuned.
`gradientn.scale`	A number, by default 0.99. It determines where the tick marks will be for the posterior probability scale. It may need to be manually tuned.
`effect.size.plt`	A numeric, by default FALSE. It determines whether the plot will color tiles by the effect size (when TRUE) or by statistical significance (when FALSE) of the augmented data including both the observed and missing patients. Assumes events are in the first column.
`extreme`	a numeric, either 0, 1, or 2. 1 leads to a standard prior specification. 0 shrinks towards 0, 2 shrinks towards 1/2. Experimental–please do not change.

Returns a list with two entries. The first FI, shows the LTFU-aware fragility index for each supplied entry of q, The last row shows the largest possible q and the corresponding LTFU-aware fragility index. The second 'info“, shows the outcome modifications that make each LTFU-aware fragility index possible.

1
2
3

mat <- matrix(nrow=2, byrow=TRUE, c(315, (1198+314)-315, 350, (1189+328)-350))
ltfu <- c(24+3, 27+5)
out <- ltfu.fi(mat, ltfu, function(m) fisher.test(m)$p.value)