Fig. 6
From: Semisupervised regression in latent structure networks on unknown manifolds
Scatterplot indicating that the responses and the 1-dimensional raw-stress embeddings are linked via a simple linear regression model. From 6-dimensional estimates of the latent positions corresponding to 100 Kenyon Cell neurons forming a directed network in larval Drosophila, 1-dimensional embeddings \({\hat{z}}_i\)’s are obtained by raw-stress minimization of the shortest path distances. The distance (\(y_i\)) between bundle entry point of the i-th neuron and mushroom body neuropil is treated as the response corresponding to the i-th neuron. Scatterplot of \((y_i,{\hat{z}}_i)\), with fitted regression line \(y=4356.1+1296.6x\) indicates a significant effect (\(p<0.01\) for \(H_0:a=0\) vs \(H_1:a \ne 0\) in \(y_i=a+b_i {\hat{z}}_i+ \eta _i\))