Skip to main content
Figure 2 | Systematic Reviews

Figure 2

From: Recovering the raw data behind a non-parametric survival curve

Figure 2

Cumulative event rates and estimated numbers at risk. (Left) Cumulative event rates in atrial fibrillation patients who received warfarin or rivaroxaban. (Right) The vertical location of each dot represents the estimated number at risk in the warfarin arm in the risk set in question (horizontal location). The numbers were derived by applying Equation 1 to the S(t j ) estimates derived from the PostScript commands used to render the vector image. The diamonds represent numbers at risk at days 0, (120), 840, reported at the bottom of the figure in the article. Clearly, even if they had not been provided, they could have been very accurately estimated just from the successive S(t j ) estimates alone. The slight lack of monotonicity in series (a) reflects rounding errors in the PostScript co-ordinates. Each n j in series (b) is based on the (clearly false) assumption that the corresponding d j =1; at these distinct failure times, clearly, d j =2, so each n j is twice that shown. Likewise, the n j s in series (c) are based on assuming d j =1, when, again clearly, d j =3, and the n j should be three times that shown.

Back to article page