3
$\begingroup$

I have 6 subjects. Each one has provided a sample every three months over a twelve month period ( 4 samples in total).

I want to account for the baseline level of each subject when I analyse the changes over time. I have been told that I need to normalise each of the subjects data points based on their baseline level, so their baseline level is ineffectively 0 when I am comparing the changes over time to this point.

Please can somebody advise how I can do this? ( Current BSc student).

Thank you

$\endgroup$
1
  • $\begingroup$ Normally, you would simply subtract out the first value from all the values (for that subject). $\endgroup$ Jul 15, 2020 at 19:12

1 Answer 1

2
$\begingroup$

To do what you are asking you simply subtract each subject's baseline measurement from all their measurements including their baseline. This will ensure that all subjects have zero baseline.

However that is a bad idea.

The main problem is that you have no way to account for possible dependency on the measurement trajectory of the baseline values. Perhaps subjects with a low baseline have a steeper trajectory. Things like that.

The first thing to do is plot your data (mesurements vs time. If you have seperate groups (eg male/female or treatment/control) then plot the groups seperately. This is to deternmine whether a linear association is plausible.

A more approprate way to model the data is a linear mixed effects model where you fit random intercepts for subjects and have time as a fixed effect. You can also allow time to vary by subject by fitting random intercepts. In R it would look something like:

model <- lmer(measure ~ time + (1|subject), data = mydata)

Ideally you would want more than 6 subects, but it's not an ideal world and 6 is a common rule of thumb to use as a minimum

$\endgroup$
1
  • $\begingroup$ Does this answer your question ? If so please consider marking it as the accepted answer. If not please let us know why $\endgroup$ Jul 31, 2020 at 4:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.