Ronald R. Halverson

Group Size and Measures of Group-Level Properties: An Examination of Eta-Squared and ICC Values

The eta-squared (η2) from a one-way random effects ANOVA is an index commonly used to estimate group-level properties of data in multilevel research. Under some circumstances, however, η2 values provide biased estimates of the group-level properties. Biased estimates occur because the magnitude of the 1.12 in a one-way rando m effects ANOVA is partially a function of group size. In this paper, the relationship between group size and η is described, and a simulation verifying the relationship between group size and 112 is conducted. The simulation demonstrates the conditions under which η does and does not provide a biased estimate of group-level properties. The paper concludes by (a) discussing corrections for η2, and (b) providing guidelines for calculating estimates of group-level properties in samples having unequal group sizes.

Benchmarking multilevel methods in leadership: The articles, the model, and the data set

Multilevel data-analytic techniques are rarely simultaneously employed and directly contrasted with each other. In this special issue of The Leadership Quarterly, hierarchical linear models (HLM), within-and-between analysis (WABA), and random group resampling (RGR) are compared and contrasted by testing the hypothesis that leadership moderates the relationship between stressors and well-being—a hypothesis that has important practical implications for the U.S. Army. This first article plays the groundwork for subsequent comparisons by testing for moderating effects using data collected from 2042 U.S. Army soldiers deployed to Haiti in November and December of 1994. Raw-score or individual-level analyses failed to find evidence of moderating effects. However, a preliminary set of group-level analyses indicated that the data had significant group-level properties that had not been modeled in the individual-level analyses. The group-level properties of the data set the stage for the three multilevel data-analytic approaches (HLM, WABA, and RGR) that are employed in three articles that follow and that are then compared and contrasted in the final article of this special issue.

Using Random Group Resampling in multilevel research: An example of the buffering effects of leadership climate☆

In this article, we provide a detailed example of how Random Group Resampling (RGR) can be used to empirically identify group effects with an example involving the buffering effects of leadership climate. RGR provides a tool for statistically determining whether group-level relationships are the result of true group phenomenon (group effects) or the result of aggregating individual data to the group level (grouping effects). Here, we present a group-level model of the stress-buffering hypothesis. Using this group-level perspective, we propose that the average perceptions of leadership climate within Army Companies will moderate the relationship between unit task significance and unit hostility. An unweighted group-means analysis revealed significant buffering effects. Following the unweighted group-means analysis, we used RGR to determine whether the significant interaction was a function of the aggregation process (grouping effects) or a function of the group-level properties of the data (group effects). The RGR analysis indicated that the interaction was related to the group-level properties or the data, and was not merely a by-product of the aggregation process. We conclude by discussing the flexibility of RGR and use it to supplement a within and between analyses (WABAs).

Cohesion and Readiness in Gender-Integrated Combat Service Support Units: The Impact of Acceptance of Women and Gender Ratio

Cohesion, combat readiness and acceptance of women were examined among male and female junior enlisted soldiers and male noncommissioned officers (NCOs) in 19 combat service support companies. The proportion of junior enlisted females in each company was negatively correlated with mean cohesion and readiness scores for junior enlisted males. The proportion of NCO females was significantly correlated with the proportion of soldiers who said they did not expect to deploy with their units, which in turn was negatively correlated with cohesion for male NCOs. For junior enlisted males, results indicated that cohesion and combat readiness increased with increased acceptance of women, but decreased as the proportion of females in the unit increased. Furthermore, acceptance of women decreased as the proportion of females in the unit increased. The results are interpreted in the light of two competing hypotheses regarding minority proportional representation-the tokenism hypothesis and the minority-proportion discrimination hypothesis.

Within- and between-entity analyses in multilevel research: A leadership example using single level analyses and boundary conditions (MRA).

This article demonstrates the use of within- and between-entity analysis (WABA) to investigate the common data set used in this special issue (N=2042 individual soldiers and J=49 Army companies). The purpose of this paper is to illustrate traditional WABA techniques and extend their use in terms of “boundary condition” analysis using multiple relationship analysis (MRA) so as to compare and contrast their results with the other techniques highlighted in this issue. Given a loose interpretation of WABA I, the results indicate that entire companies reporting a positive leadership climate had low levels of psychological hostility. However, the relationship between task significance and psychological hostility appears conditional upon leadership climate. Under poor leadership conditions, the level of task significance in the unit as a whole was negatively related to levels of psychological hostility. In contrast, under good leadership conditions, a lower level of analysis appears to be operative: Individual reports of task significance were negatively related to psychological hostility.

Integrating multilevel analyses and occupational stress theory

In this chapter, we integrate occupational stress theory with emerging analytic and theoretical considerations related to multilevel modeling. We begin by differentiating among models at different levels, and identify the inferential errors that can inadvertently arise when applying occupational stress findings to organizations. Second, we discuss the basic framework for using multilevel modeling to study occupational stress processes over time. Finally, we apply the implications of the first two sections to a popular occupational stress model. In so doing, we show how multilevel theory and methodology can be used to enhance our understanding of occupational stress processes. The conclusion of this chapter is that multilevel theory and analytic techniques have much to offer occupational stress researchers from both a theoretical and methodological perspective.

Determinants of Soldier Support for Operation Uphold Democracy

A substantial number of U.S. Army soldiers deployed to Haiti for Operation Uphold Democracy did not believe it was important that the U.S. military be involved in the operation (49%); did not believe that what the U.S. military was doing was important (38%); and did not believe in the overall value of the operation (43%). At the same time, a substantial number of soldiers had positive feelings about what they were doing in Haiti and the mission they were accomplishing. The primary focus of this investigation was to examine factors that were related to the wide variation in soldier reports of support for Operation Uphold Democracy. The results indicated that a combination of soldier characteristics (e.g., race, gender), unit characteristics (e.g., unit type), task characteristics (e.g., task significance), and operational characteristics (e.g., perceptions of public support) accounted for nearly 50% of the variance in soldier reports of support for the overall operation.