Help with interpreting a cumulative probability distribution graph in a journal paper

Question

I am from the life sciences field (neuroscience research) and I am reading a journal paper about the formation of dendritic spines. In their experiments, they count the number of protrusions along dendrites and compare them across the different experimental conditions.

In one of the figures of this journal paper, they show some cumulative probability distribution graphs:

I have read online that with a cumulative probability plot, one can read off the probability of being above or below a particular value, or of being within, or outside, a particular range. However, I am having some trouble interpreting these graphs (Figures C and F) in the context of the paper.

For Figure C, would I be correct in saying that the probability of having greater than 20 protrusions/50 μm of dendrite is greater in the Insulin group compared to the Vehicle group? This is because in the Insulin group, you have a ~60% probability that you will have greater than 20 protrusions/50 μm of dendrite. Whereas in the Vehicle group, you have a ~20% probability that you will have greater than 20 protrusions/50 μm of dendrite.

I am also wondering, given raw data how do people make these cumulative probability distribution plots, and what value can they give to the reader? Any advice is appreciated.

Answer 1

1. The "cumulative probability" listed on the y-axis of, e.g., plot C is the proportion of the distribution that is less than or equal to the corresponding value on the x-axis. For 20 protrusions per 50 μm, 60% of the insulin distribution was below that threshold, whereas 85% (?) of the vehicle distribution was below that. More generally, for plot C, the insulin distribution (solid circles) is always to the right of the vehicle distribution (hollow circles). That shows that insulin tends to have higher values.
2. Be careful with "the probability of having greater than 20 protrusions/50 μm of dendrite is greater in the Insulin group compared to the Vehicle group", as this sounds like a claim based on a statistical test, whereas the plot should be considered descriptive. (FWIW, I would be surprised if it isn't significant, though.)
3. Presumably most any statistical software could make such a plot. You would sort your data and compute the proportion of your data from the minimum value to the maximum value for each unique value (of, e.g., the number of dendrites). Then you would plot a line (or line of connected circles) for these sorted pairs. It may be common to show this plot in neuroscience, but I think it would be more appropriate to make an empirical cumulative distribution plot.
4. As a first pass for less familiar readers, boxplots will show much the same information, but be easier to understand. The value of the CDF plot is that you can see the whole distribution (assuming readers will be able to interpret it fluently).