Please enable JavaScript.

Coggle requires JavaScript to display documents.

Klein From Micro to Meso: Critical Steps in Conceptualizing and …

- - - - 2 should Faust simply conduct
        his individual-level research as planned, analyze the results, and generalize them to the
        team level of analysis?
      - 1 Should Faust then collect his individual-level data as planned and then simply
        aggregate the data to the team level, conducting team-level analyses of the data?
      - , 3 should
        Faust simply conduct individual-level analyses, after first collecting individual-level
        measures of the predictors (e.g., leader charisma), aggregating these measures to the
        team level, and then assigning the aggregated (average) scores back to individual team
        members such that the predictors are identical for the members of each team?
- - - - the relationships between team location (a global construct) and within-team homogeneity of team efficacy (a configural construct). #
  - - - Depending on the nature of the team task, team performance may reflect the following: the sum of individual team members’ contributions (the better each individual performs, the better team performance); the poorest team member’s contribution
        (the team can perform no better than its weakest performer);
        
        Possible operationalizations, as we have
        hinted, include the sum of individual team member values, indices of variability
        among team member values, the minimum or maximum value among a team’s members, and measures of the team network (e.g., density, homophyly).
        
        In sum, individual-level data are critical for measuring both
        configural and shared team properties, but only shared team properties require the
        demonstration of within-group consensus or consistency
- - - - A cross-level, direct effects model suggests that a predictor variable at one level of analysis influences an outcome variable at a different—typically lower—level of analysis.
        
        Note that cross-level models of this type can only explain between-unit (here, between-team) variability in individual-level outcome measures.
        
        Faust might predict that team leader charismatic leadership (a group-level, shared construct) has a direct effect on team member self-efficacy (an individual-level outcome variable)
      - Cross-level moderator models suggest that variables at two different levels of analysis (e.g., one group-level variable and one individual-level variable) interact to predict an outcome at the lower level of analysis (e.g., an individual-level outcome measure).
      - The third type of cross-level model—thecross-level frog-pond model describes the effects of individual group member’s standing within a group on individual-level outcomes
        
        Frog-pond models suggest, in essence, that
        the true or absolute size of a frog is irrelevant.
        
        Faust might propose that an individual’s performance relative to his or her team members’ performance predicts individual self-efficacy. Thus, for example, an individual whose absolute, true level of performance is mediocre may have high self-esteem if he or she works with others who perform relatively poorly.
- - - - Shared experiences and interactions within a team may foster such within-team homogeneity and between-team variability. In contrast, a strong organizational culture and consistently applied organization-wide policies may foster within-organization homogeneity—that is, homogeneity within and across teams
      - However, cross-level models do require special considerations, as they typically rest on an assumption of within-unit homogeneity and between-unit variability
        on some unit-level constructs and an assumption of within and between-unit variability (that is, between-individual variability within and across units) on other, individual-level constructs.
      - for example, that Faust predicts that team leader charisma (a shared team construct) moderates the relationship of self-efficacy and performance (two individual-level constructs). The ideal sample in which to test this cross-level moderator model is one in which (a) team leader charisma varies betweenbut not within-teams, and (b) both self-efficacy and performance vary both within- and between-teams. Cross-level frog-pond models also require both within-unit and between-unit variability on the predictor, frog-pond variable
      - The sampling implications of multilevel homologous models are complex, in part because such models are more common in theory than in empirical research. Indeed,
        one can find few—if any—empirical examples of homologous models. One strategy to test such a model is to collect a different sample of data to represent each level of the homologous mode
- - - - Justifying the Aggregation of Lower Level
        Data to Higher Units of Analysis
        
        Each of the procedures commonly used to justify aggregation provides an assessment of the extent to which lower level data (e.g., individual-level data) are homogeneous within units (e.g., groups), as assumed of shared unit-level constructs.
        
        Thus, Faust might
        use any or all of these procedures to assess the extent to which a shared team constructs—say, team efficacy—is indeed homogeneous, or shared, within teams.
        
        r (wg) index compares the variability of a given variable within a specific unit (here, team) to an expected variance (EV). If the variability within a unit is substantially smaller than the variability expected by chance, then the resultingr wg value suggests that it is justifiable to aggregate lower level data, for the specific variable and specific unit in question, to the unit level of analysis.
        
        Eta-squared. The eta-squared value from a one-way analysis of variance provides an estimate of the extent to which individual-level variability on a given measure is explained by higher level units.
        
        Bliese and Halverson (1998) have criticized the use of
        eta-squared, demonstrating that eta-squared provides inflated estimates of betweenunit variability especially when group sizes are small (less than 25 people per group)
        
        The intraclass correlation coefficient (1) is calculated
        using the results of a one-way analysis of variance (i.e., eta-squared) and is commonly interpreted in much the same way as eta-squared.
        
        Intraclass correlation (2). The intraclass correlation (2), or ICC(2), answers the following question: How reliable are the group means within a sample? ICC(2) values, like other measures of reliability (e.g., Cronbach’s al ha), are commonly considered acceptable if they equal or exceed .70. athematically, ICC(2) is a variation of ICC(1). Essentially ICC(2) is a function of ICC(1) adjusted for group size.
        
        WABA. Like ICC(1) and ICC(2), WABA builds on eta-squared, and resulting estimates of within-unit and between-unit variability.
        
        (а) эта мера показывает достаточное внутриблочное соглашение и вариабельность между единицами, чтобы оправдать агрегирование меры и использовать ее для введения в действие конструкции общей единицы (целостности); или (б) эта мера демонстрирует достаточно изменчивость внутри единицы и достаточное межблочное соглашение для обоснования использования меры для введения в действие переменной (части) лягушки-пруда; или
        (c) мера варьируется как между единицами, так и внутри единиц и, следовательно, должна использоваться для введения в действие отдельных уровней конструкций (двусмысленных).
        
        Thus, Faust might find between-unit variability in team efficacy
        to be both practically and statistically significant, neither practically and statistically
        significant, or either practically or statistically significant but not both. Faust would
        use WABA I to assess within- and between-group variability in all of the measures of
        his model, not just team efficacy.
        
        Пример объясняющий Waba
        
        In unambiguous cases—when a measure’s variance
        1 lies quite predominantly between-units,
        2 or quite predominantly within-units,
        3 or roughly equally within- and between-units : :!:
      - Analyzing the Relationships Among Variables:
        Testing Single-Level Relationships,
        Multilevel Homologous Relationships,
        and Cross-Level Relationships
        
        WABA. As suggested above, WABA I, is designed to assess the extent of within-unit and between-unit variance in a single measure. WABA II is designed to assess the extent to which two or more variables covary primarily within-units, between-units, or both within- and between-units.
        
        Thus, if Faust used WABA I and II, he would analyze all of his data, examining the extent to which all of the measures vary and covary between- and within-teams.
        
        WABA is not designed to test cross-level direct effect models, thus Faust could not test the simultaneous effects of, say, team efficacy (wholes) and self-efficacy (equivocal) on performance (equivocal). Rather, Faust would use WABA to test single-level relationships in the data.
        
        CLOP encompass a number of analysis of variance, covariance, and regression approaches designed to examine the main or direct effects of higher level (e.g., team) variables on lower level (e.g., individual) outcomes and/or the moderating effects of higher level variables on lower level relationships
        
        CLOP answers the following questions: What is the effect of a higher level unit characteristic on a lower level outcome? and/or, What is the influence of a higher level unit characteristics on the relationship between lower level variables?
        
        for example, Faust wanted to assess the effects of team location (a group-level, global construct) and team leader charisma (a group-level, shared construct) on team member performance (an individual-level outcome variable)
        
        Like CLOP, HLM is designed to test cross-level direct effect and moderating effects models. Thus, HLM—like CLOP—answers these important questions: What is the effect of a higher level unit characteristic on a lower level outcome? And/or, What is he influence of a higher level unit characteristic on the relationship between lower level variables? Furthermore, like CLOP, HLM includes no procedures to justify aggregation.
        
        CLOP and HLM may be used, as we have suggested, to test the same models. #
        
        Thus WABA differs from CLOP and HLM in two essential respects. First, WABA includes procedures and standards to evaluate the level of variables and of elationships (including procedures and standards to justify the aggregation of lower level variables to higher levels of analysis): CLOP and HLM do not include such procedures. And second, WABA is typically used to assess single-level models, although it can also be used to test multilevel homologous models: CLOP and HLM, in contrast, are typically used to assess cross-level relationships, although clearly they can also assess single-level models.