Hello John, Was I not clear? See below. In a message dated 10/29/2011 1:01:29 P.M. Central Daylight Time, John.Antonakis@UNIL.CH writes: Hi George:My suggestions are not misleading (at least methodologically not so); perhaps what I said was not clear enough. Let me try explain things again, citing a few more papers.Here are two key points:1. Whether leaders learn or not from their subordinates depends, in part, on how smart leaders are (there are other reasons too why leaders will learn, e.g., the characteristics of the tasks, which includes getting feedback to be able to learn, Shanteau, 1992; subordinate competence, firm-level factors, etc.). The question is how supervisors learn from their subordinates, not how much they learn as I read it. No matter how they test on IQ tests, if they do not listen properly(Graen,2012), they learn very little. In fact, many sources of information can be closely monitored by an interested supervisor. The use of such sources by supervisors and real leaders can be measured (Ask Bill Schiemann). He a SIOP F Now, to establish whether there is a causal effect of x on y, the predictor/s of y must be exogenous; that is, estimating y = b0 + b1x + e must satisfy the assumption of the orthogonality of e to x. If x correlates with e, which it will if x is endogenous, then b1 cannot be consistently (correctly) estimated. The only way to get a consistent effect is to find an "instrument"--and exogenous source of variance--that will purge x from endogeneity bias. Suppose that "x" is what Ronit will measure (whether a leader learns from subordinates); the problem is that x may correlate with e, so the correct specification to have is to use an instrumental-variable regression. Intelligence could serve as an instrument: if it is exogenous it means that it is stochastic with respect to unmodeled sources of variance that predict y (i.e., e). I submit that this will not establish causality. This procedure cannot eliminate possible plausible alternative explanations and establish the proper order (See Popper). I sum yours is not a shortcut. See quantum Physics. Use of the General Linear Model does not establish causality only proper research design will do the work. Cheers George Graen
Hello John, Was I not clear? See below. In a message dated 10/29/2011 1:01:29 P.M. Central Daylight Time, John.Antonakis@UNIL.CH writes: Hi George:My suggestions are not misleading (at least methodologically not so); perhaps what I said was not clear enough. Let me try explain things again, citing a few more papers.Here are two key points:1. Whether leaders learn or not from their subordinates depends, in part, on how smart leaders are (there are other reasons too why leaders will learn, e.g., the characteristics of the tasks, which includes getting feedback to be able to learn, Shanteau, 1992; subordinate competence, firm-level factors, etc.). The question is how supervisors learn from their subordinates, not how much they learn as I read it. No matter how they test on IQ tests, if they do not listen properly(Graen,2012), they learn very little. In fact, many sources of information can be closely monitored by an interested supervisor. The use of such sources by supervisors and real leaders can be measured (Ask Bill Schiemann). He a SIOP F Now, to establish whether there is a causal effect of x on y, the predictor/s of y must be exogenous; that is, estimating y = b0 + b1x + e must satisfy the assumption of the orthogonality of e to x. If x correlates with e, which it will if x is endogenous, then b1 cannot be consistently (correctly) estimated. The only way to get a consistent effect is to find an "instrument"--and exogenous source of variance--that will purge x from endogeneity bias. Suppose that "x" is what Ronit will measure (whether a leader learns from subordinates); the problem is that x may correlate with e, so the correct specification to have is to use an instrumental-variable regression. Intelligence could serve as an instrument: if it is exogenous it means that it is stochastic with respect to unmodeled sources of variance that predict y (i.e., e). I submit that this will not establish causality. This procedure cannot eliminate possible plausible alternative explanations and establish the proper order (See Popper). I sum yours is not a shortcut. See quantum Physics.
Hi George:My suggestions are not misleading (at least methodologically not so); perhaps what I said was not clear enough. Let me try explain things again, citing a few more papers.Here are two key points:1. Whether leaders learn or not from their subordinates depends, in part, on how smart leaders are (there are other reasons too why leaders will learn, e.g., the characteristics of the tasks, which includes getting feedback to be able to learn, Shanteau, 1992; subordinate competence, firm-level factors, etc.).
Now, to establish whether there is a causal effect of x on y, the predictor/s of y must be exogenous; that is, estimating y = b0 + b1x + e must satisfy the assumption of the orthogonality of e to x. If x correlates with e, which it will if x is endogenous, then b1 cannot be consistently (correctly) estimated. The only way to get a consistent effect is to find an "instrument"--and exogenous source of variance--that will purge x from endogeneity bias. Suppose that "x" is what Ronit will measure (whether a leader learns from subordinates); the problem is that x may correlate with e, so the correct specification to have is to use an instrumental-variable regression. Intelligence could serve as an instrument: if it is exogenous it means that it is stochastic with respect to unmodeled sources of variance that predict y (i.e., e).