In Noa's paper "The effect of individual variation on the structure and function of interaction networks in harvester ants", she had only one network. She inferred individual variation in degree from the shape of the degree distribution--small world, normal, exponential, poisson, etc.
I can't do this because individuals' degrees change over time. When I plot multiple time series (spaghetti plot), I want to know whether there's a consistent difference between the lines, or whether they aren't distinguishable, statistically.
Noa suggests using repeatability analysis, similar to how researchers in animal personality quantify whether personality is consistent over time. She says to use the ICC, or "intraclass correlation coefficient" to do this.
In the current version of the model, individuals do not have different degree preferences or tendencies. Differences in degree emerge from the random assortment of the model, but as we can see from this plot, these differences are not consistent and only persist as long as the memory of the model (2 time steps).
Intuition seems to be a good guide here. Can look at this graph and state pretty confidently that these individuals do not differ significantly in their degree. But I can't quantify that statistically.
Once I update the model to include individual variation in degree, I will want to compute the ICC for that model to see if we can actually distinguish individuals, and compare the results to this (null-ish) model.
In Noa's paper "The effect of individual variation on the structure and function of interaction networks in harvester ants", she had only one network. She inferred individual variation in degree from the shape of the degree distribution--small world, normal, exponential, poisson, etc.
I can't do this because individuals' degrees change over time. When I plot multiple time series (spaghetti plot), I want to know whether there's a consistent difference between the lines, or whether they aren't distinguishable, statistically.
Noa suggests using repeatability analysis, similar to how researchers in animal personality quantify whether personality is consistent over time. She says to use the ICC, or "intraclass correlation coefficient" to do this.
For example, see https://www.datanovia.com/en/lessons/intraclass-correlation-coefficient-in-r/#:~:text=The%20Intraclass%20Correlation%20Coefficient%20(ICC,with%20two%20or%20more%20raters.
Use:
Intuition seems to be a good guide here. Can look at this graph and state pretty confidently that these individuals do not differ significantly in their degree. But I can't quantify that statistically.
Once I update the model to include individual variation in degree, I will want to compute the ICC for that model to see if we can actually distinguish individuals, and compare the results to this (null-ish) model.