In both models, univariate analyses were used to assess the relationship between HIV status, receipt of HAART, contraceptive use, and covariates.
Before logistic regression modeling, univariate analyses between the outcome (participation to a cluster) and each variable were independently performed using the Fisher's exact test to select potentially interesting variables to include in the logistic model.
The linear model univariate regression analyses were performed before (Model 1) and after (Model 2) adjusting for relevant covariates (age, sex, BMI, smoking and hypertension).
To evaluate the uncertainty of parameter values and the robustness of the model, univariate sensitivity analyses were performed for each parameter in the model over the ranges.
To evaluate the uncertainty of parameter values and the robustness of the model, univariate sensitivity analyses were performed for each parameter in the model over the ranges shown in Tables 1, 2, 3. Ranges were sourced from reported literatures, such as 95CI% %, or a range±20% of the base case value when reported data were not available.
Separate General Linear Model (GLM) Univariate analyses were conducted to identify the reward related (Reward-Proximity and I-Won) and loss related (I-Lost and Near-Miss) neurons.
The variables were included in logistic regression models using univariate analyses to establish the likelihood of association with the improvement of VA with RE correction.
The multiple logistic regression models in univariate analyses were used to examine which factors had independent effects on patients in full remission as compared with those in persistent depression, and patients in full remission as compared with those in episode.
In each of these 100 bootstrap samples, we repeated all modeling steps described above for deriving the initial model (including the univariate analyses, subsequent variable selection, and graphical examination of continuous variables for potential transformation and imposition of thresholds) to derive a new risk score and estimates of outcome incidence in each stratum.
We will include factors in the model which in univariate analyses have a p-value < 0.10.
