摘要:

Data for Toledo, Ohio and San Antonio, Texas reveal that a decline in the neonatal mortality relationship has led to a slight narrowing of the overall inverse relationship between infant mortality and SES in both cities. At the same time, there are notable differences in both the magnitude and pattern of the neonatal changes. Certain characteristics of the two cities are cited as possible explanations of the observed differences.

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1035

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP

摘要:

MacMillan and Daft (a) have argued that ratio and component regressions dealing with relationships between organization size and administration yield the same inferences about administrative economies, diseconomies, or returns to scale, presumably removing much of the current confusion traced to the “definitional dependency” argument. Unfortunately, their claim is unwarranted because their mathematical proofs are inadequate, ignoring the correlation between component variables A and P, and the simulated data are highly restricted in the A, P relationship. Our simulated data, with varying error variance and 100 samples for each condition, show considerable inference inconsistency among the several analytical procedures. We suggest continued use of the A, P regression intercept as a test of the scale condition as long as the slope of the regression is significant.

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1040

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP

摘要:

Feinberg and Trotta report simulation experiments of administrative scale relationships that show inference inconsistency across statistical models. Their findings are contrasted with our previous analysis (MacMillan and Daft, a) that reported inference consistency across statistical models with respect to definitional dependency. This paper reconciles the two sets of findings by concluding that definitional dependency is not a problem in ratio variable research, but measurement error may cause inference inconsistency. Three recommendations are suggested for investigators who use ratio variables to study organization scale and similar sociological relationships.

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1059

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1068

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP

摘要:

MacMillan and Daft (b) argued from a limited number of computer-simulated samples that results of the correlations between ratios of the type E/P and N/P are consistent with the implications of the linear relationships between E and P and N and P. P designates measurement of population size. We used a considerably larger number of samples, simulated according to their specifications, and we found that their argument is unwarranted. We examined their argument under other circumstances as well, to find the effects of variation in the simple correlations among the component variables and of differences in the way particular concepts might be measured. We concluded in all the circumstances that we cannot rely on ratio correlations to yield the same conclusions as implied by what is “known” about the data.

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1076

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1087

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1089

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1091

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1092

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP

ISSN:0430-0696    

DOI:10.1093/sf/62.4.1095

出版商:The University of North Carolina Press    

年代:1984

数据来源: OUP