China
Africa
Explore chapters and articles related to this topic
A Review of Herbivore Effects on Seaweed Invasions
Published in S. J. Hawkins, A. J. Evans, A. C. Dale, L. B. Firth, D. J. Hughes, I. P. Smith, Oceanography and Marine Biology, 2017
Enge Swantje, Sagerman Josefin, Sofia A. WikströM, Pavia Henrik
The meta-analysis on consumer preference for non-native versus native seaweeds was conducted using the metafor-package in R (Viechtbauer 2010) and the OpenMee software (Dietz et al. 2016). The weighted overall mean effect of herbivore preference for non-native or native seaweeds was calculated by a random-effects model using the restricted maximum-likelihood estimator for residual heterogeneity. Bootstrapped 95% confidence intervals were calculated for the overall mean effect size generated from 4999 iterations. To check the robustness of the meta-analysis outcome, we calculated the fail-safe number with the weighted method of Rosenberg (2005), which represents the number of additional studies with no effect needed to change the result of the meta-analysis from significant to non-significant. Publication bias was further examined with a funnel-plot and the rank correlation test for funnel plot asymmetry (Begg & Mazumdar 1994). The influence of outliers on the overall mean effects size was tested by evaluating the change of the overall effect when one study at a time was left out of the analysis. Since hypothesis-driven research tends to favour large effect sizes in support of the hypothesis in earlier publications, we examined temporal trends in the data with a cumulative meta-analysis sorted by publication year (Jennions & Møller 2002).
BMDP
Published in Paul W. Ross, The Handbook of Software for Engineers and Scientists, 2018
3V—General Mixed Model ANOVA Maximum likelihood and restricted maximum likelihood for fixed and random effects modelsEstimates of parameters, asymptotic standard errors, t statisticsLog-likelihoods and likelihood ratio test
Policy evaluations of the Renewable Fuel Standard and agricultural land use changes in three Midwestern states for a decade: variables that influence significant changes in crop decisions are not all about ethanol
Published in Biofuels, 2023
Krista Russell, Nicholas Guehlstorf
To test each of the three hypotheses, a time series mixed-model ANOVA with an unstructured variance–covariance structure model was analyzed in PROC MIXED of SAS (version 9.4). The repeated measures interval within the model was one year, or the equivalent of one growing season. The subject was each county nested hierarchically within its given state. Model effects were estimated using restricted maximum likelihood (REML), as this method allows for the use of the necessary multivariate variance–covariance structure. The final model was: where Y is the vector containing the response variable; β is the vector of parameter coefficients; and X is representative of the design matrix containing the values for the independent variables. Independent variables include: state, latitude, longitude, the previous year’s acreage devoted to the particular crop under study, percent change in CRP, crop price (corn price or soybean price, model depending), the average number of days between planting and R1 for a given county and year, the median precipitation during the planting season for a given county and year, and the median precipitation during harvest season for a given county and year. In corn, the R1 phase of crop development is when silks become visible outside of the corn husk, and in soybeans, R1 is when the plant starts to flower. The vectors u and ε contain the random error terms associated with the errors between and within counties, respectively.
Adaptations in driver behaviour characteristics during control transitions from full-range Adaptive Cruise Control to manual driving: an on-road study
Published in Transportmetrica A: Transport Science, 2020
Silvia F. Varotto, Haneen Farah, Klaus Bogenberger, Bart van Arem, Serge P. Hoogendoorn
The ‘Mixed Model’ command in SPSS 24 (IBM Corporation 2016) was used for model estimation. The estimation method chosen was the restricted maximum-likelihood (REML), which provides unbiased estimators of the variance components accounting for the degrees of freedom used to estimate the fixed effects (Verbeke and Molenberghs 2009; Zuur et al. 2009). The parameters estimated were used to calculate the marginal means of the driver behaviour characteristics over time in each traffic conditions controlling for the system state, between-subjects variation and residual covariance structure. Pairwise comparisons were used to test statistically the hypothesis of significant changes in the mean driver behaviour characteristics over time when drivers are in control of the vehicle (I or AAc) in different traffic flow conditions. Mean values at time t were compared to mean values at time t+ 1. Significant changes in each second over a certain interval of time after the ACC system was deactivated or overruled by pressing the gas pedal can be interpreted as an indicator of the time duration needed to stabilise driving behaviour after resuming manual control (transition period, similar to Merat et al. [2014]). The magnitude of the corresponding adaptation in driver behaviour characteristics was calculated using the model. The advantage of this data analysis technique is to quantify the transition period explicitly based on significant changes in the driver behaviour characteristics. The final results are robust to the initial choice of the 20-s time interval for each transition.
Longitudinal MRI data analysis in presence of measurement error but absence of replicates
Published in IISE Transactions on Healthcare Systems Engineering, 2018
Chitta Ranjan, Kamran Paynabar, Martin Reuter, Kourosh Jafari-Khouzani, the ADNI
Unlike MLE, the restricted maximum likelihood (REML) approach yields smaller bias, and hence, is typically, used for LME parameter estimation. In REML, we construct a likelihood that depends only on θ to remove the effect of degrees of freedom lost in the estimation of β. We use a Bayesian formulation where the fixed effect, β, and random effect, b's, are considered as random variables. We take a flat prior distribution for β, p(β) ∼ N(β*, Γ), Γ → ∞, which is a non-informative density and the choice of β* is immaterial. The prior for b is chosen as p(b) ∼ N(0, Q).