I finally found a study that demonstrates a correlation between fasting insulin levels and weight loss. In this study, they manipulated fasting insulin levels through macronutrient composition an diets administered at various caloric levels. Below is the scatter plot of the data for all subjects:
Horizontal axis = change in fasting insulin level between time points
Vertical axis = change in weight
See the correlation?
Neither do I.
So many studies these days begin with correlative analyses of various variables obtained in various studies. But if the investigator sees a scatter plot like above - a blind person throwing darts at a target might yield a tighter cluster! - they move on. Because there really is nothing to see here!
It is bad enough that if the plot had looked more like the one below, one would see all manner of causation assigned to the correlation seen in the scatter plot below.
At least we see a first quadrant trend towards a positive correlation (increase in horizontal axis → corresponds to increase on the vertical axis ↑). But when there's not even a correlation to be seen on the scatterplot from my study. It's so "all over the map" as to require +/- on both axes, or all 4 quadrants with a smattering of points to be had in each.
Yes. This is the Grey & Kipnis data re-plotted (roughly estimated from poor quality plots but makes no difference!) .