Can longitudinal generalized estimating equation models distinguish network influence and homophily? An agent-based modeling approach to measurement characteristics.

BACKGROUND: Connected individuals (or nodes) in a network are more likely to be similar than two randomly selected nodes due to homophily and/or network influence. Distinguishing between these two influences is an important goal in network analysis, and generalized estimating equation (GEE) analyses of longitudinal dyadic network data are an attractive approach. It is not known to what extent such regressions can accurately extract underlying data generating processes. Therefore our primary objective is to determine to what extent, and under what conditions, does the GEE-approach recreate the actual dynamics in an agent-based model. METHODS: We generated simulated cohorts with pre-specified network characteristics and attachments in both static and dynamic networks, and we varied the presence of homophily and network influence. We then used statistical regression and examined the GEE model performance in each cohort to determine whether the model was able to detect the presence of homophily and network influence. RESULTS: In cohorts with both static and dynamic networks, we find that the GEE models have excellent sensitivity and reasonable specificity for determining the presence or absence of network influence, but little ability to distinguish whether or not homophily is present. CONCLUSIONS: The GEE models are a valuable tool to examine for the presence of network influence in longitudinal data, but are quite limited with respect to homophily.

Multi-scale heterogeneity in vegetation and soil carbon in exurban residential land of southeastern Michigan, USA.

Exurban residential land (one housing unit per 0.2-16.2 ha) is growing in importance as a human-dominated land use. Carbon storage in the soils and vegetation of exurban land is poorly known, as are the effects on C storage of choices made by developers and residents. We studied C storage in exurban yards in southeastern Michigan, USA, across a range of parcel sizes and different types of neighborhoods. We divided each residential parcel into ecological zones (EZ) characterized by vegetation, soil, and human behavior such as mowing, irrigation, and raking. We found a heterogeneous mixture of trees and shrubs, turfgrasses, mulched gardens, old-field vegetation, and impervious surfaces. The most extensive zone type was turfgrass with sparse woody vegetation (mean 26% of parcel area), followed by dense woody vegetation (mean 21% of parcel area). Areas of turfgrass with sparse woody vegetation had trees in larger size classes (> 50 cm dbh) than did areas of dense woody vegetation. Using aerial photointerpretation, we scaled up C storage to neighborhoods. Varying C storage by neighborhood type resulted from differences in impervious area (8-26% of parcel area) and area of dense woody vegetation (11-28%). Averaged and multiplied across areas in differing neighborhood types, exurban residential land contained 5240 ± 865 g C/m2 in vegetation, highly sensitive to large trees, and 13 800 ± 1290 g C/m2 in soils (based on a combined sampling and modeling approach). These contents are greater than for agricultural land in the region, but lower than for mature forest stands. Compared with mature forests, exurban land contained more shrubs and less downed woody debris and it had similar tree size-class distributions up to 40 cm dbh but far fewer trees in larger size classes. If the trees continue to grow, exurban residential land could sequester additional C for decades. Patterns and processes of C storage in exurban residential land were driven by land management practices that affect soil and vegetation, reflecting the choices of designers, developers, and residents. This study provides an example of human-mediated C storage in a coupled human-natural system.

Exploring agent-level calculations of risk and returns in relation to observed land-use changes in the US Great Plains, 1870-1940.

Land-use change in the U.S. Great Plains since agricultural settlement in the second half of the nineteenth century has been well documented. While aggregate historical trends are easily tracked, the decision-making of individual farmers is difficult to reconstruct. We use an agent-based model to tell the history of the settlement of the West by simulating farm-level agricultural decision making based on historical data about prices, yields, farming costs, and environmental conditions. The empirical setting for the model is the period between 1875 and 1940 in two townships in Kansas, one in the shortgrass region and the other in the mixed grass region. Annual historical data on yields and prices determine profitability of various land uses and thereby inform decision-making, in conjunction with the farmer's previous experience and randomly assigned levels of risk aversion. Results illustrating the level of agreement between model output and unique and detailed household-level records of historical land use and farm size suggest that economic behavior and natural endowments account for land change processes to some degree, but are incomplete. Discrepancies are examined to identify missing processes through model experiments, in which we adjust input and output prices, crop yields, agent memory, and risk aversion. These analyses demonstrate how agent-based modeling can be a useful laboratory for thinking about social and economic behavior in the past.