A practical and example filled tour of simple and multiple regression techniques (linear, logistic, and Cox PH) for estimation, adjustment and prediction.

Loading...

From the course by Johns Hopkins University

Statistical Reasoning for Public Health 2: Regression Methods

81 оценки

A practical and example filled tour of simple and multiple regression techniques (linear, logistic, and Cox PH) for estimation, adjustment and prediction.

From the lesson

Module 2B: Effect Modification (Interaction

Effect modification (Interaction), unlike confounding, is a phenomenon of "nature" and cannot be controlled by study design choice. However, it can be investigated in a manner similar to that of confounding. This set of lectures will define and give examples of effect modification, and compare and contrast it with confounding.

- John McGready, PhD, MSAssociate Scientist, Biostatistics

Bloomberg School of Public Health

Greetings, and welcome to lecture set five.

In this lecture set, we're going to talk about another phenomena that

involves looking at a two variable relationship outcome and

exposure, and incorporating information about a behind the scenes third variable.

But, unlike the previous situation with confounding,

we're not necessarily concerned about whether this behind the scenes variable is

distorting the overall outcome exposure relationship.

But, we're concerned about whether or not we should be estimating separate outcome

exposure relationships for different levels of this variable.

So, for example,

suppose we do a randomized clinical trial to look at the association of a treatment

on the risk of relapse in cancer patients compared to the standard of care.

The style is randomized, so we're not necessarily concerned about

the overall association between treatment and relapse being confounded by

other factors related to whether or not the patient got the treatment, like sex.

However, we still might be concerned as to whether the association between

the relapse in treatment is the same for males and females.

So, the question is not whether or not that one overall association of

interests is being distorted by differences in

the sex distributions between the two groups in this particular example.

It wouldn't be, because of the randomization, but the question is

about whether or not we should have one overall outcome exposure relationship for

both sexes together.

Or whether or not there is evidence that the relationship between the outcome

exposure, relapse, and treatment differs for men and woman.

Perhaps, the drug works well for women, and not so well for men.

So, this idea of effect modification, or

statistical interaction, is sometimes a question of interest in a research study.

So, in this section here,

we're going to talk about investigating effect modification.

Look at some examples of where it was investigated in a research context.

And, sometimes found, and sometimes not found.

And, we're going to work to distinguish it from what we talked about in

the previous lecture set, confounding.

And, then further down the line, we'll show how to use multiple

regression techniques to officially investigate effect modification, i,e,

a differing outcome exposure relationship for

different levels of a third factor in the context of multiple regression.

Coursera делает лучшее в мире образование доступным каждому, предлагая онлайн-курсы от ведущих университетов и организаций.