Fig. 2
From: Semisupervised regression in latent structure networks on unknown manifolds
Plot showing consistency of the substitute estimator of the regression parameter vector on known manifold. For 100 Monte Carlo samples, substitute estimates are computed using the projections of the optimally rotated adjacency spectral estimates of the latent positions onto the manifold, and then the sample MSEs of the estimator based on the true regressors and the substitute estimator are computed. For graphs of moderate size (\(n \le 2000\)), the substitute estimator performs significantly worse than the estimator based on the true regressors. However, as the number of nodes increases, the difference in performances of the estimators diminish and the mean squared errors of both the estimators approach zero