From da2a39f33e9cc20cfab02ec9b0625940ab5fcdd0 Mon Sep 17 00:00:00 2001 From: Mitch Macdonald <35157616+mcmacdonald@users.noreply.github.com> Date: Sun, 12 Jan 2025 21:07:55 -0700 Subject: [PATCH] README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 2496b7a..789a469 100644 --- a/README.md +++ b/README.md @@ -10,7 +10,7 @@ where p(*k*) is the probability distribution function (PDF) of degree *k*, *k* r The pk() command calculates the gradient of the slope and standard errors that provide the probable range of the slope. The script uses ordinary least squares to regress the logged PDF of the degree distribution on the logged degree *k*. The log-log regression implies that the gradient of the slope scales by kY orders of magnitude. A negative slope ≥ -2 indicates that the network is centralized and scale-free i.e., there is greater degree disassortativity (heterophily), such that majority of small degree nodes attach themselves to the minority of large degree nodes in the upper-tail of the degree distribution. The steeper the slope, the more centralized the network is. A slope that is < 2, by comparison, indicates that the network is more decentralized and not scale-free i.e., there is greater degree assortivitiy (homophily). As the slope → 0, the more decentralized the network is. -Further, this function includes adjusments to the calculation of the scaling exponent based on the critique of this model by Clauset et al. (2009). The first is that I use the procedures in Gabaix & Ibragimov (2011) to calculate the standard error of the slope [Clauset et al. (2009) explain why ordinary least squares does not accurately calculate the standard error of the power-law slope]. The second is that the function provides the option to use the cumulative distribution function (CDF) of the degree distribution, rather than the probability distribution function (PDF) of the degree distribution to caculate the model parameters [Again, Clauset et al. (2009) explain why the CDF provides better estimates of the scaling exponent than the PDF]. +Further, this function includes adjusments to the calculation of the scaling exponent based on the critique of this model by Clauset et al. (2009). The first is that I use the procedures in Gabaix & Ibragimov (2011) to calculate the standard error of the slope [Clauset et al. (2009) explain why ordinary least squares does not accurately calculate the standard error of the power-law slope]. The second is that the function provides the option to use the cumulative distribution function (CDF) of the degree distribution, rather than the probability distribution function (PDF) of the degree distribution, to caculate the model parameters [Again, Clauset et al. (2009) explain why the CDF provides better estimates of the scaling exponent than the PDF]. I illustrate the procedures on different types of social networks. The original data is published by The Mitchell Centre for Social Network Analysis, University of Manchester [https://sites.google.com/site/ucinetsoftware/datasets/covert-networks].