Explore chapters and articles related to this topic
Detection of Cerebrovascular Changes Using Magnetic Resonance Angiography
Published in Ayman El-Baz, Jasjit S. Suri, Cardiovascular Imaging and Image Analysis, 2018
Yitzhak Gebru, Guruprasad Giridharan, Mohammed Ghazal, Ali Mahmoud, Ahmed Shalaby, Ayman El-Baz
Statistical analysis was performed using R software, version 3.3. A mixed effects linear model was used to test the relationship of MRA data with clinical BP measurements. Brain slices were separated into upper (above circle of Willis) and lower (below circle of Willis) compartments to determine correlation with clinical BP readings. The circle of Willis, near the brain base, is where the intracranial cerebral arteries take off from and give rise to progressively smaller vessels [5]. The BP measurements were combined into a single value, the estimated mean arterial pressure MAP=(2×DBP+SBP)/3, which was a covariate in the model. Also included in the model were patient age, gender, and a random intercept per patient. The dependent variable was the mean of the Euclidean distance map over the entire vascular tree within each compartment. (Two separate models were fit to the upper and lower compartments.) Statistical significance of fixed effects in the fitted models was determined using likelihood ratio chi-square tests.
Design of Experiments and Its Deployment in SAS and R
Published in Tanya Kolosova, Samuel Berestizhevsky, Supervised Machine Learning, 2020
Tanya Kolosova, Samuel Berestizhevsky
In general, fixed-effect parameters are associated with one or more continuous or categorical parameters and describe the relationships of the covariates to the dependent variable for an entire population. Fixed effects are unknown constant parameters, and estimation of these parameters is of fundamental interest, as they describe the relationships of the covariates with the continuous outcome variable. In experimental settings, for fixed-effect factors data are gathered from all the levels of the factor that are of interest.
Crossover and Repeated Measures Designs
Published in John Lawson, Design and Analysis of Experiments with R, 2014
The output on the previous page shows that there is no significant group effect or carryover difference, and no significant direct treatment effect. The between subjects variance component, σs2, is estimated to be 1.5078, while the within subject variance component, σ2, is estimated to be 4.563. The Fixed effects: produces results that are equivalent to the Anova F-tests.
Re-Examining the Effect of Audience Response Systems on Learning Outcomes: Evidence from the Last Decade
Published in International Journal of Human–Computer Interaction, 2023
Tuncer Akbay, Neşe Sevim-Cirak, Osman Erol
In this study, a random-effects model rather than the fixed-effect model was utilized because the fixed-effects model assumes that there is only one true effect size that all included studies try to find, whereas random effects model assumes that true effect size may vary from study to study (Borenstein et al., 2007). Due to many circumstances (e.g., demographic characteristics of the subjects in studies, the number of participants in studies, the intensity of the intervention in the experimental studies, more reliable measurement of the effect, etc.), the true effect may vary from one study to another. Field and Gillett (2010), in fact, suggested using random-effects model in a meta-analysis conducted in social sciences. Hence, it would be appropriate to use the random effects model when conducting meta-analysis with the research findings in these situations. The results of the meta-analysis presented in the following sections.
Autism, Attachment, and Alexithymia: Investigating Emoji Comprehension
Published in International Journal of Human–Computer Interaction, 2022
Hannah Taylor, Christopher J. Hand, Hannah Howman, Ruth Filik
To determine the individual and combined effects of emoji type, AQ, TAS, attachment anxiety (ANX) and attachment avoidance (AVO) on classification accuracy, we generated a series of linear mixed effects models in R (R Development Core Team; http://www.r-project.org). Across participants and trials, there were 7,740 data points available for analysis. We used the “lme4” package (Bates et al., 2015); we followed a generalized linear mixed-effects approach using the “glmer” command and added the argument “family = binomial,” given the nature of our classification data. Optimal random effect structures were identified using forward model selection (see Barr et al., 2013; Matuschek et al., 2017). The random effect structure for these models included only random intercepts by participants and items (more-complicated error terms resulted in non-convergence in full and reduced models). Fixed effects were tested using likelihood-ratio tests comparing full and reduced models. Post-hoc tests were conducted using the “emmeans” package (v1.4.8, 26/06/20; Lenth et al., 2020), and significance thresholds adjusted using the Bonferroni method. An observed power analysis conducted using the PowerSim function of the “simr” package in R (Green & MacLeod, 2016) determined that given our sample size and number of observations, our analyses were 100% fully powered.
Effects of social information signals on user engagement: evidence from randomized field experiments
Published in Behaviour & Information Technology, 2022
Swanand J. Deodhar, Ayushi Tandon, Abhas Tandon, Abhinav Tripathi
Our estimation models for CC (Equation 1) and CO actions (Equation 2) are presented below. In these equations, i indexes the user, w indexes the word, and d indexes the wordlist. The coefficients represented by β0 in each equation represent the influence of social information signal on the user i's respective engagement actions for the word w. The rest of the symbols carry the following meanings: the term α captures the user-fixed effects, and the term δ accounts for wordlist fixed effects. Using these multiple fixed effects, over and above the randomisation, controls for different sources of unobserved heterogeneity. Lastly, ϵ captures the error term.