Researchers often use a two-step statistical process, involving univariate and multivariate regression analysis, to see if a certain variable (like age, BMI, etc.) is associated with an outcome (like live birth rate).
In a univariate regression analysis, only one variable is considered. This is kind of like a screening process to see which variables are associated with the outcome, before moving onto the multivariate analysis that looks at the big picture.
For example, a study might look at how a number of variables impact live birth rates (like age, BMI and AMH).
In a univariate analysis, they might find that age and BMI are associated, but not AMH.
So now that AMH is screened out, we can focus on age and BMI.
But what if age and BMI are connected and they don’t independently affect live birth? For example, what if older people generally have a higher BMI? In this case, is it the older age itself, or the higher BMI itself, that’s affecting live birth? Since they’re connected it’s hard to know (because someone who’s older will also have a high BMI).
To isolate this, you can use a multivariate regression analysis to see if age and BMI independently affect live birth. This analysis evaluates the independent effects of multiple variables (e.g., age and BMI) on the outcome simultaneously.
After a multivariate analysis, we might find that only age independently affects live birth, and not BMI.
After a univariate and multivariate regression, you can see which variables are statistically significant and independently affect the outcome. It also lets you control, or statistically adjust, for other variables to see what the independent effects are.