The present article aims to (1) conceptualize and present the two-level multilevel model for e-collaboration research, (2) conceptualize the Intra-Class Correlation Coefficient (ICC), (3) conceptualize R2 in e-collaboration multilevel modeling, (4) present centering methods that can be used in e-collaboration multilevel modeling, (5) present parameter estimation and hypothesis testing methods for e-collaboration multilevel modeling, and (6) list some of the existing commercial software packages that can be used for analyzing the e-collaboration multilevel data.
Key Terms in this Chapter
Intra-Class Correlation (ICC): It is the proportion of the total variance explained by the group differences.
Amount of Variance Explained at Level-2 in each Regression Coefficient (R2): Is an index of the amount of variance explained in each of the Level-1 random regression coefficients (intercept and slopes) at Level-2 by the q Level-2 predictors. Here, R2 is estimated by comparing the Level-2 variance (t00 or tpp) estimate from the basic two-level model with the Level-2 variance, t00, estimate from the no-predictor model. (See Equation 8 and 9).
t00: Is the Level-2 (group-level) variance among the random intercepts,ß0j, from the Individual-Level (Level-1) model.
t11: Is the Group-Level (Level-2) variance among the random slopes, ßpj, from the Individual-Level (Level-1) model.
Multilevel Modeling: Refers to a set of different statistical methods for analyzing data with two or more levels of hierarchical or nested structures.
Centering: Is changing the location of the Level-1 and Level-2 predictors in the multilevel model by subtracting the grand mean, group mean, or specific value from each predictor variable.
Two-Level Multilevel Model: Refers to analytic methods for data with two levels where individuals are nested within e-collaboration groups and there are predictors characterizing individuals as well as groups.
Amount of Variance Explained at Level-1 (R2): Is an index of the amount of variance explained at Level-1 by the p Level-1 predictors. Here, R2 is estimated by comparing the Level-1 variance, s2, estimate from the basic two-level model with the Level-1 variance, s2, estimate from the no-predictor model. (See Equation 7)
Empirical Bayes Estimate of a Random Level-1 Coefficient: Is an estimate of the Individual-Level random intercept or slope(s) for a particular e-collaboration group. This empirical Bayes estimate utilizes the data from that specific e-collaboration group and the data from all the e-collaboration groups.