Σύγκριση μεθόδων

	Quadratic Assignment Procedure	MRQAP Network Regression
Πεδίο	Sociology	Sociology
Οικογένεια	Process / pipeline	Regression model
Έτος προέλευσης	1976 (QAP); 1988 (network application)	1988 (MRQAP); 2007 (double-semipartialing test)
Δημιουργός	Lawrence Hubert & James Schultz; David Krackhardt	David Krackhardt; David Dekker, David Krackhardt & Tom Snijders
Τύπος	Permutation-based test of association between two matrices	Permutation-based multiple regression for dyadic (matrix) outcomes
Θεμελιώδης πηγή	Krackhardt, D. (1988). Predicting with networks: Nonparametric multiple regression analysis of dyadic data. Social Networks, 10(4), 359–381.	Krackhardt, D. (1988). Predicting with networks: Nonparametric multiple regression analysis of dyadic data. Social Networks, 10(4), 359–381.
Εναλλακτικές ονομασίες	QAP correlation, QAP permutation test, matrix permutation test, Hubert-Schultz QAP	MRQAP, multiple regression QAP, Dekker double-semipartialing, QAP regression
Συναφείς	4	4
Σύνοψη	The quadratic assignment procedure (QAP) is a permutation-based method for testing the association between two relational matrices measured on the same set of actors — for example, whether who advises whom is correlated with who is friends with whom. Because the dyads in a network are not independent, ordinary correlation and regression give invalid p-values; QAP fixes this by comparing the observed matrix correlation to a reference distribution generated by randomly relabeling the nodes of one matrix many times.	Multiple regression quadratic assignment procedure (MRQAP) extends QAP to the regression setting: it predicts a dependent relational matrix from several independent relational matrices on the same actors — for example, modeling who collaborates with whom as a function of who is co-located, who shares a department, and who has prior friendship. Coefficients are estimated by ordinary least squares on the vectorized matrices, but significance is assessed by permutation, because dyadic dependence invalidates the standard regression standard errors.