Suchergebnisse

Artifacts contained in random data may either strengthen or weaken the evidence for a res'er's hyp. A recent paper by John Shelton Reed ('A Note on the Control of Lynching,' Public Opinion Quarterly, 1969, 33, 268-271), provides an example of an artifact arising because both the describing variable & the response variable have an element in common. The artifact in this case favors the hypothesis of the paper. It is shown by means of equations & a random number analysis that Reed's evidence is too weak to prove his hyp. Some alternate approaches which would diminish the bias from the artifact are suggested. The most reasonable approach in this case would be to collect monthly data, instead of the use of overlapping yearly data. John Shelton Reed (U of North Carolina, Chapel Hill), in RE- PLY TO TUFTE, states that E. R. Tufte's observation that Reed's r is at least partly tautological is 'quite correct.' However, Tufte is unduly harsh with Reed's conclusion. Reed replicated Tufte's experiment on 40 runs of 21 paired comparisons & achieved results which were not too far from his original conclusion. Disagreement is voiced with Tufte's statement that 'the best way to approach the matter is to use regression.' The nonparametric test seems preferable in this application, notably because several difficulties involving auto-? of the series have been resolved for it. These difficulties are not addressed at all by the regression analysis proposed by Tufte. This, however, is a minor point. The interested reader should go to the original presentation of the test. M. Maxfield.

Random responses can be attributed to a series of variables which are not functionally related to the content of the instrument; the effects of such random responses on the testing of substantive hypotheses are investigated. R's were separated into 2 groups (on the basis of an item for which all change was random): stayers, who did not change over time; & movers, who changed randomly. To determine the degree to which relationships between variables at any single point in time could be strengthened by identifying & removing the 'movers', identical analyses were carried out for each group separately. After separating movers & stayers, the corrections between responses to 2 measures of political efficacy were examined for all 3 waves of a panel study. Considerable variability was found in the correlations, but correlations were consistently higher for the stayers. The greatest constraint between the 2 items was shown by those who stayed over all 3 waves (the most stable group). Thus, the relationship between the 2 measures of efficacy was considerably strengthened by removing the movers. The index of political saliency was found to be the most predictive variable for both mover groups & the whole sample; the age of 1st party identification was most closely related to voting punctuality for all 3 stayer groups. The mover-stayer distinction is pertinent in inferring the predictive power of specific independent variables. 2 Tables. H. Dorian.

Social scientists are unaware that the absence of a relationship between an independent & dependent variable is just as spurious as its presence. The main reason is that the relationship is concealed by a suppressor test factor. The logical status of suppressor test factors & compensating influences are considered. Examples of antecedent, intervening, & component suppressor variables are provided. Though formally the same, standard test factors & suppressor test factors generate opposite interpretations of spuriousness in a causal sense. Several substantive contributions of suppressor variables to social science are discussed. Suppressor variables may appear as often as standard test factors in research. 4 Tables. S. Karganovic.

One motivation for using panel data is to determine the relative effects of two varibles which are known to be related. Here are reviewed the four techniques that have been used most often to address this problem. The relative merits of the four -- (1) Lazarsfeld's 16-Fold Table, (2) Coleman's Four-State Continuous Time Markov Process, (3) Cross-Lagged Panel Correlation, & (4) Path Analysis -- & the particular problems associated with each are reviewed. The first three methods are shown to yield similar results, the fourth to do so under some assumptions. 11 Tables, 3 Figures. AA.

DATA ON POLICY EXPENDITURES FOR 23 STATES THAT EXPERIENCED POLITICAL & ECONOMIC CHANGE DURING THE DECADE FROM 1956 TO 1966 ARE USED TO TEST THE POWER OF BOTH LINEAR & NONADDITIVE MODELS IN EXPLAINING VARIANCE IN EXPENDITURES. DATA ON EXPENDITURES FOR HIGHWAYS, EDUCATION, WELFARE, & HEALTH WERE USED. THE INDEPENDENT VARIABLES WERE URBANIZATION, INDUSTRIALIZATION, MEAN FAMILY INCOME, & MEAN SCHOOL YEARS COMPLETED; POLITICAL VARIABLES WERE LEGISLATIVE COMPETITION, GUBERNATORIAL COMPETITION, LEGISLATIVE APPORTIONMENT, & VOTER PARTICIPATION. THE NONADDITIVE MODEL ASSUMES THAT STATE POLICY-MAKING IS INCREMENTAL UNTIL A PLATEAU IS REACHED WHEN POLITICAL ISSUES WILL RESULT IN PROPORTIONATELY HIGHER EXPENDITURES FOR SERVICES. FOR WELFARE EXPENDITURES BOTH MODELS REFLECT EXPLAINED VARIANCE EQUALLY. FOR THE REST, THE NONADDITIVE MODEL EXPLAINS MORE VARIANCE. THE NONADDITIVE MODEL EXPLAINS MORE VARIANCE IN ALL POLICY EXPENDITURES OVER THE DECADE. IT IS A MORE ACCURATE REFLECTION OF THE DYNAMICS OF POLICY OUTPUT BECAUSE IT INDICATES THE IMPORTANCE OF POLITICAL VARIABLES. 2 FIGURES, 4 TABLES. D. BURKENROAD.

The European Commission : Top Managers and "Management Culture"

If status puts the Commission officials in a position susceptible of being interpreted according to national belonging, practice shows that the different styles can combine and define a european practice. The policy project, the diversity of jobs, and the multicultural nature of the environment stimulate the individuals but they sometimes feel destabilised by the variableness of management styles and cultural references. The alternative use of the concepts of "management" and of "hierarchy" frames daily work, "management" which bumps into the "conflict of national cultures" would have as objective the reinforcing of professional culture. The question is to know whether "management" serves to reduce cultural differences or to develop their respective contributions.

Entropy analysis is suggested as a useful technique for the examination of multivariate relationships among categorical variables. It makes no assumptions about the numerical properties of variables nor about the nature of the interrelationships among sets of variables. The technique is able to describe completely the explanatory power of all the constraints within a set of variables & to partition them into the following categories: distributional constraints, internal constraints, external constraints, & conditional entropy (unconstrained). Following a discussion of the entropy concept & outlining the statistics, an entropy analysis is performed on P. Converse's data ("The Nature of Belief Systems in Mass Publics" in Apter, D. [Ed], Ideology and Discontent, New York: Free Press, 1974), controlling for social groups. The results show that the responses of the mass sample to the several political issue items are as constrained as those of the candidate sample. 5 Tables. Modified HA.

Central to the investigation of the potential facilitative or inhibitive effects of distraction upon persuasion is a conceptualization of distraction that allows for an adequate measurement of the distraction variable. To avoid bias distraction must be measured in terms of the extent to which a person's reception of a distracting stimulus is increased &/or his reception of the original stimulus is decreased. A formula is devised to avoid favoring one theory over another & to account for S's' differences in information handling ability. Such a formula also possesses the property which enables all measurements of distraction to fall between 0 & 1, a useful property for many statistical analyses. 1 Table. AA.

A strategy for detecting the interaction effect of independent variables recently was proposed by J. N. Morgan & J. A. Sonquist (see SA 0104/E0777). The approach utilizes the AID program & requires the analyst to determine if interaction is present between 2 variables by visually inspecting a plot of average values for each category of each variable. The assumption was that no statistical test was available which would serve this purpose. This note suggests that Tukey's Test for Additivity can be used to statistically test the signif of interaction between the 2 variables. AA.

A demonstration of the use of probit analysis, a statistical technique which circumvents the ways in which dichotomous dependent variables violate the assumptions of the classical regression model which are crucial to the interpretation of the estimates produced by the regression technique. As an illustration, probit analysis is applied to survey data to produce some valuable measures of the effect of precinct level campaigning in a municipal election. The probit model is applied to data estimating the effect of precinct-level personal contact campaigning (door-to-door canvass & telephone solicitation) on voter behavior. The data analyzed come from a sample of the 3 competitive City Council districts from Rochester, NY's 4 districts in 1967. The results indicate that personal contact campaigning in a municipal election has substantially the same effect as it does in nat'l elections. M. Duke.