Hi James M. Clark Professor of Psychology 204-786-9757 204-774-4134 Fax j.cl...@uwinnipeg.ca Department of Psychology University of Winnipeg Winnipeg, Manitoba R3B 2E9 CANADA
>>> <sbl...@ubishops.ca> 07-Dec-09 10:11 AM >>> ... But here's the catch. They provided a complex statistical analysis (to me, anyway) but their analysis depends on a curious grouping of birth order: first-borns comprised one group, and later-borns the other. But the later-born group also included only children (without siblings). On logical grounds, one would think that only children belong in the first-born category instead. Their justification for doing this was inspection of the data. For trust: "Means of x [their monetary datum] for middleborn, lastborn and only children appeared much closer to each other than to the mean of x for firstborns (Table 2); these three categories were therefore pooled." For reciprocity: "Only children and laterborns were pooled because their average amounts sent (y) were closer to each other than to the average amount sent by firstborns (Table 2)." My own inspection of their data suggests that without this post- hoc categorization, they would not have been able to report significant results. Is their move kosher, or do we have a case of data-massaging here? JC: Stats were nonconventional (randomization tests), but looks to me like they got a significant effect (.042) WITHOUT the grouping (i.e., using the 4 groups First, Middle, Last, Only) and then grouped them to show that the significant variability was due to First vs Non-First (other 3 groups). Depending on hypothesized underlying mechanism (I did not read rest of paper, just results), it could make sense to group Only with non-First born since, for example, they would have no younger siblings. Of course that reasoning would not apply to middle born, who were also lumped together and actually showed results most different from Firstborns. Take care Jim --- To make changes to your subscription contact: Bill Southerly (bsouthe...@frostburg.edu)
