What test and post tests should I use?

#1
I have rats that received a 30 minute infusion of control (saline) followed by 30 minutes of a low, mid, and high dose (total 120 minutes). A measurement (heart rate) was taken once during each dose level and control for a total of 4 timepoints. One group of 8 rats received this regiment under one type of anesthesia and another group of twelve received the regiment after a different anesthetic. I want to take means and compare the anesthetics at each matched time point as well as compare each timepoint against the control timepoint for that particular anesthetic. What tests should I run?
 

Karabiner

TS Contributor
#2
I want to take means and compare the anesthetics at each matched time point as well as compare each timepoint against the control timepoint for that particular anesthetic.
Why are you doing thise? What is the actual research question which you want to answer?

With kind regards

Karabiner
 
#3
How each different anesthetic affects heart rate. Different dose levels were applied to see how each anesthetic responded to the challenge.
 

Karabiner

TS Contributor
#4
How each different anesthetic affects heart rate. Different dose levels were applied to see how each anesthetic responded to the challenge.
So why don't you conduct a repeated mesaures ANOVA, with time of measurement as repeated-measures factor (4 Levels) and groups as between-subjects factor (2 levels)? Seems like you are looking for the Time x Group interaction.

With kind regards

Karabiner
 
#5
Thanks, when I brought this up to a statistician they mentioned you need 3 or more groups for anova. In addition, if a couple values are missing and RM is not an option what do you suggest?
 

hlsmith

Not a robit
#6
R.tylor,

Correct, A better modeling strategy is to use multi-level modeling (where readings are clustered in rats). This model also runs when you have missing data. Though, traditionally this is a more difficult and sophisticated model that isn't always easy for new-comers to run.

Tell us more about your missing data and the underlying mechanism.
 
#7
There is not a lot of data missing...one rat was missing from 2 time points.

I ran this study because in many cases it is best to evaluate the effect of certain agents on heart rate in anesthetized animals because this provides a more stable model (no variations as a result of temperament in conscious animals). I know one is better in that heart rate changes are not confounded by the depression caused by anesthetics, but of course for publication I will need to describe the differences between the two anesthetics. I just can’t figure out the best methods of statistically comparing/describing the difference between the two anesthetics and how they differ in the way the heart responds to treated conditions (the ascending dose of an agent that increases heart rate).

Thanks!
 

hlsmith

Not a robit
#8
8 vs 12 with missing data plus needing interaction terms in the model doesn't bode well for analytics unless there is a large treatment effect.
 
#11
Correct, so you would get a coefficient for increase over time (slope) for both groups and see if they were disordinal (not parallel and potentially crossing). Given the sample size I would imagine imputation would be out.