—
title: “Example 7”
author: “sheng huo”
date: “5/11/2017”
output: html_document
—
“`{r setup, include=FALSE}
library(knitr)
setwd(“C:/Users/NOAH/Desktop/CSC 495/week7/ex7/ex7”)
read_chunk(“example07.R”)
knitr::opts_chunk$set(echo = TRUE)
“`
## QAP tests
In this example, we work with the second Lord of the Rings network.
### Step 1: Load the necessary libraries
Note that SAND must come after network.
“`{r C1, results=”hide”, warning=FALSE, message=FALSE}
“`
### Step 2: Load the data and summarize
“`{r C2}
“`
### Step 3: Load the CUG and QAP utilities
“`{r C3}
“`
### Step 4: Reminder: CUG test for LOTR3 / Race
If we use “Race” feature as is, we will get an error because of zero values.
“`{r C4}
“`
### Step 5: Reminder QAP test for LOTR3 / Race
“`{r C5}
“`
### Step 6: Showing results
“`{r C6}
“`
## ERGM Example
We are looking at three networks.
### Load networks
“`{r C14}
“`
### Plotting
“`{r C15}
“`
### Triad census
Directed networks so we can look at the triad census
“`{r C16}
### Comparing metrics
Assortativity by color, density, transitivity, reciprocity
“`{r C17}
“`
### Plotting
Use melt to get a ggplot-friend graph.
Note: assortativity rising, density decreasing, transitivity / reciprocity not really changing
“`{r C18}
“`
### Convert to network
ERGM only works on network objects. The `intergraph` package provides the conversions.
“`{r C19}
“`
### Run some basic models
“`{r C20}
“`
### Load pre-computed models.
_Do this when your model fitting will take more than a minute or so._ You do not want to have to wait every time you generate the HTML.
“`{r C21}
“`
### Compare AIC of different models
“`{r C22}
“`
### Summaries of 3 and 6
“`{r C23}
“`
### Tweaking model 6
“`{r C24}
“`
### Fitting networks 4 and 5
“`{r C25}
“`
### Interpreting coefficients
Need the inverse logit function.
A Red->Blue edge is more than 20x more likely at time 3 than time 5. Mutual edges are equally likely in both times.
“`{r C26}
“`