Meta-analyses combine different studies for overall estimates which are viewed as stronger evidence than any single study. It is important to assess the potential for selection bias.
The complex sample design in NHANES enables analysts to properly estimate statistics pertaining to the United States population.
Many of our projects include comparisons of group means while checking for interaction and adjusting for potential confounders.
Receiver operating characteristic analysis measures how well a new diagnostic test or risk model discriminates between patients with and without a “gold standard” diagnosis or event.
Decision trees can break a population down according to risk to simply identify groups likely to benefit from preventive treatment.
Logistic regression provides equations estimating the odds of subjects developing into an incident case of disease over some specified time interval.
Cox regression is frequently used for risk models assessing the likelihood of disease-free survival.
Group comparisons of changes in repeated or continuous measures may be performed with the use of general linear mixed models.
Statistically powerful generalized estimating equations use multiple measures for each subject to provide definitive comparisons, in this case with only six pigs in four groups.
We have also used weighted linear regression and correlation analysis among many other methods selected on a project specific basis.