I am wondering if it would help to show the same points for predictive quantiles in panel (C) as are shown in the other two panels? This might help clarify that indeed we start with the same quantiles as the input to the ensembling process. Then maybe you could use... like, black squares or something, at the (common) points where the cdfs are evaluated and averaged to get the linear pool?
I am suggesting this because at first when I saw the figure, I was thinking we were demonstrating that it is natural to compute a linear pool based on cdf value inputs, but the text in the caption is describing the process for computing a linear pool based on quantile inputs.
I am wondering if it would help to show the same points for predictive quantiles in panel (C) as are shown in the other two panels? This might help clarify that indeed we start with the same quantiles as the input to the ensembling process. Then maybe you could use... like, black squares or something, at the (common) points where the cdfs are evaluated and averaged to get the linear pool?
I am suggesting this because at first when I saw the figure, I was thinking we were demonstrating that it is natural to compute a linear pool based on cdf value inputs, but the text in the caption is describing the process for computing a linear pool based on quantile inputs.