Skip to content

Latest commit

 

History

History
84 lines (61 loc) · 2.01 KB

classify-passfail-analysis.md

File metadata and controls

84 lines (61 loc) · 2.01 KB

R markdown for Classifying student passing and failing from network measures

Adrienne Traxler & Jesper Bruun 9/3/2020

Summarize results here.

Update 6/9: Started with collecting data information (since I keep forgetting which variables to use).

Summary of data

Networks

After the network data has been cleaned and assigned node attributes (by loadAllNetworks.r), and centrality values are calculated (by calculatePR_TE_H.r), the igraph objects are stored in data/PRTEH.RData.

That file includes:

  • accPS, accCD, and accICS: Accumulated network objects for each week for the problem solving, concept discussion, and in-class social network layers.
  • accPS_PR, etc.: The accXX_Y-named objects hold centrality calculations for each layer for PageRank, Target Entropy, Hide. For TE and Hide, these are lists of vectors (one per week). For PageRank, it’s a list of lists (one per week), with the centrality values we want in the vector entry of the list.

As part of standardizing the networks, they all have the same number of nodes, 166.

sapply(accPS, vcount)
## [1] 166 166 166 166 166 166 166

sapply(accCD, vcount)
## [1] 166 166 166 166 166 166 166

sapply(accICS, vcount)
## [1] 166 166 166 166 166 166 166

Node information

There are also data frames of node information (after make_node_data_frames.Rmd), in data/centrality_data_frames.Rdata. This has:

  • dfPS, dfCD, dfICS: A data frame for each layer listing week, node name, all predictor variables, and pass/justpass outcomes.
  • nPass and nJustPass: Tables of pass/fail counts (coded 1/0), for everyone and for people who were right on the border. For nJustPass, people not on the border are coded NA.
load("data/centrality_data_frames.Rdata")
nPass
## 
##   0   1 
##  38 128

nJustPass
## 
##    0    1 <NA> 
##   28   39   99

Logisitic regression results