Scale Space Methods for Analysis of Type 2 Diabetes Patients' Blood Glucose Values

<table>Top: a family plot of the density estimate for the times of BGC readings for the full cohort over the full project period. Each line is a Gaussian kernel smooth with increasing bandwidth. The thick line corresponds to the data-driven bandwidth, <svg height="10.95" id="M68" style="vertical-align:-0.1254pt" version="1.1" viewbox="0 0 24.1875 10.95" width="24.1875" xmlns="http://www.w3.org/2000/svg">
<g transform="matrix(1.25,0,0,-1.25,0,10.95)">
<g transform="translate(72,-63.24)">
<text transform="matrix(1,0,0,-1,-71.95,63.41)">
<tspan style="font-size: 12.50px; " x="0" y="0">ℎ</tspan>
<tspan style="font-size: 12.50px; " x="10.677562" y="0">=</tspan>
</text>
</g>
</g>
</svg>8.7 days.  Bottom: a SiZer plot of the data in the upper figure. In shades from light to dark these correspond to (i) significant negative gradient, (ii) no significant gradient, (iii) not enough data for inference (does not appear in this particular map), and (iv) significant positive gradient. The horizontal line corresponds to the data-driven bandwidth which is displayed in bold in the upper panel. The dashed white lines visualize four times the bandwidth at each scale.</table>

Computational and Mathematical Methods in Medicine

fig2

Figure 2

Figure 2: Scale Space Methods for Analysis of Type 2 Diabetes Patients' Blood Glucose Values 