I have two groups as control and study. I measured blood stem cell levels in each groups. I compared the levels with t-test between two groups and found statistical significance between two groups. However, i got a critique by a reviewer that age and sex may be a confounding factor effecting stem cell levels. So, how can i adjust age and sex when comparing two groups.
$\begingroup$
$\endgroup$
2
-
$\begingroup$ How do you measure stem cell level - % of something ? $\endgroup$– user10619Commented Jun 7, 2017 at 13:09
-
$\begingroup$ ANCOVA is what you're looking for. It's equivalent to the regression approach described below but historically as been used for this issue. $\endgroup$– NoahCommented Jun 10, 2017 at 5:12
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
The easiest way to do this, in my opinion, is to run a regression. Your model will simply be something like "blood stem cell levels = alpha + GROUP + age + sex + error".
Group is a binary variable that you code as 1 or 0 depending on what group they are in. Sex is coded the same way.
What this does is control for age and sex. It lets you just see the affect of group.
-
$\begingroup$ Maybe it would make sense to add age*sex too in case that those features interact $\endgroup$ Commented Jun 7, 2017 at 14:13
-
$\begingroup$ You are right that this controls for age and sex, however controlling for possible confounders can open back doors to many other confounders. For example, if your two groups are rich and poor, then controlling for age might open up the confounder diet. For example, because poor people who make it to old age might have a great diet. In general controlling for confounders is not a great approach to eliminating all confounders. $\endgroup$– Neil GCommented Jun 13, 2017 at 20:12