Matched Data
Peter Cummings in Analysis of Incidence Rates, 2019
All the odds ratios in these tables are biased. To get the correct odds ratio of 0.33, we can use the mhodds command, which can compute an odds ratio stratified on the matching variable, vehid (vehicle identifier) in this example. This analysis uses only the 40 pairs in which subjects are discordant on both the exposure (seat belt use) and the outcome (died): . mhodds died belt vehid Mantel-Haenszel estimate of the odds ratio Comparing belt==1 vs. belt==0, controlling for vehid note: only 40 of the 150 strata formed in this analysis contribute information about the effect of the explanatory variable ------------------------------------------ Odds Ratio chi2(1) P>chi2 [95% Conf. Interval] ------------------------------------------ 0.333333 10.00 0.0016 0.162954 0.681857 ------------------------------------------
Potentials and Limits of Phenomenological Models
Tiziana Rancati, Claudio Fiorino in Modelling Radiotherapy Side Effects, 2019
The explanatory variables that are usually observed are items that are already recorded in the treatment process or variables that are known or expected to be potentially predictive. These include patient characteristics, such as demographic information, anatomy, risk factors, comorbidities, and baseline toxicities; disease characteristics, such as tumor location and stage; and treatment characteristics, such as treatment modality and dose distribution. Medical images of the patient provide additional or supporting information that is increasingly utilized. Before analysis this large quantity of data is usually reduced to a comprehensive set of candidate predictors. For example, the three-dimensional dose distribution is often reduced to a set of dose-volume parameters in selected organs, which are assumed to provide a biologically relevant and consistent dosimetric description.
Statistics in medical research
Douglas G. Altman in Practical Statistics for Medical Research, 1990
The way round this problem is, perhaps surprisingly, related to the technique of linear regression described in Chapter 11. I showed there how to describe the relation between two variables, or, more specifically, how the value of one variable can be predicted from the value of the other. This method too can be extended, to allow us to predict the value of a variable from the values of several other variables. In other words, we have a single dependent (outcome) variable and two or more explanatory (predictor) variables. The method is called multiple regression. The explanatory variables can be either continuous or binary or categorical. Multiple regression can thus be used to regress birth weight on sex and gestational age. It can be shown that all analysis of variance problems can also be analysed in the framework of multiple regression (see section 12.4), but for balanced data sets (usually from experiments) it is more common to keep to the analysis of variance approach.
Prevalence and factors associated with the occurrence of sexual violence among people with disabilities in Burkina Faso
Published in AIDS Care, 2022
Harouna Fomba, Henri Gautier Ouedraogo, Kadari Cissé, Seni Kouanda
Explanatory variables. Explanatory variables included in the analyses were individual, relational and contextual. Individual variables were represented by the gender (male/female), age (15–34 years old, 35–54 years old, and 55–69 years old), education levels (no education, primary school, secondary school or higher education), marital status (Single, married, divorced, or widow), income level (regular income, irregular income, or no income), alcohol consumption (yes, no), tobacco consumption (yes, no), age when the disability started (since birth, childhood, or adulthood), nature of the disability (eye, ear, mobility, mental, instinct, and/or communication). As for the relational variables, we have considered multiple sexual partner frequentation (men or women who had had more than one partner sexual in the last 12 months) (yes, no), place where the person usually spends time (at home, in the neighbourhood or outside the neighbourhood). The only contextual factor was the residence area (rural or urban).
A flexible semiparametric regression model for bimodal, asymmetric and censored data
Published in Journal of Applied Statistics, 2018
Thiago G. Ramires, Edwin M. M. Ortega, Niel Hens, Gauss M. Cordeiro, Gilberto A. Paula
In many practical applications, the response variables are affected by explanatory variables. In the presence of explanatory variables with nonlinear effects, semiparametric models are widely used and when their models provide a good fit, they tend to give more precise estimates of the quantities of interest. Recently, several regression models have been proposed in the literature by considering the class of location models. For example, Ramires et al. [18] introduced the log-beta generalized half-normal geometric regression model for censored data, Cordeiro et al. [8] presented the log-generalized Weibull-log-logistic regression model for predicting longevity of the Mediterranean fruit fly and Ortega et al. [17] studied a power series beta Weibull regression model for predicting breast carcinoma. A disadvantage of class of the location models is that the variance, skewness, bimodality, kurtosis and other parameters are not modeled explicitly in terms of the explanatory variables but implicitly through their dependence on the location parameter. As an alternative, the GAMLSS [20], wherein the systematic part of the model is expanded, allows not only the location but all parameters of the conditional distribution of Y to be modeled as parametric functions of explanatory variables.
Joint regression modeling of location and scale parameters of the skew t distribution with application in soil chemistry data
Published in Journal of Applied Statistics, 2022
F. Prataviera, A. M. Batista, P. L. Libardi, G. M. Cordeiro, E. M. M. Ortega
In many practical applications, the response variables are affected by explanatory variables. In the presence of explanatory variables with linear effects, parametric models are widely used and tend to give more precise estimates of the quantities of interest when give good fits. Several regressions have been proposed in the literature by considering the class of location models. For example, [30] extended the beta regression by considering a systematic component for the precision parameter, [22] developed a heteroscedastic nonlinear regression under scale mixtures of SN distributions, [27] introduced the log-beta generalized half-normal geometric regression for censored data, [10] presented the log-generalized Weibull-log-logistic regression for predicting longevity of the mediterranean fruit fly, and [26] introduced a new class of heteroscedastic log-exponentiated Weibull regressions.
Related Knowledge Centers
- Data Mining
- Density
- Design of Experiments
- Experiment
- Confounding
- Design of Experiments
- Reliability Engineering
- Risk Factor
- Medical Statistics
- Feature
- Omitted-Variable Bias