How do I account for both intra- and inter-individual variance when comparing the control with a conditioned test stimulus? I have data that consists of 10 repetitions of a measurement per person, for all repetitions there is one control value and one test value (after conditioning). This data will be reproduced in multiple people.
I am now thinking how to introduce the intra-individual variance into the statistical test. Without repetitions I would simply use a paired t-test if I can expect a normal distribution of the data.
I would be very thankful if someone could point me to some literature or explain to me how to analyze this sort of experiment.
Edit: Answers to Michelle's questions
This is a neurophysiological study. The goal is to observe the influence of conditioning peripheral electrical stimulation on the amplitude of reflex responses. Thus there is always a pair of reflexes first with 10 s no activity/stimulation before (control) and afterwards again 10 s no activity/stimulation followed by 10 s of conditioning.
Here is a small schematic (. no stimulation, I reflex stimulation and - conditioning stimulation):
.........I..........----------I
This paradigm is repeated per person 10 times, i.e. 20 measures per person, always one control matched to one test measure.
The time lag between the end of conditioning and test is important and will be investigated in the future but for this experiment there is only a fixed delay. Nonetheless I would be thankful to know how to handle this too.
I do not believe that there is a known mechanistic process which affects the values.
 A: Have you looked at repeated measures analysis of covariance, where the control ("baseline") reflex amplitude would be the covariate? I believe you'll need a design that incorporates the baseline as I imagine that is correlated to, if not predictive of, the post conditioning amplitude. The intra-individual variance ("within subjects variance") is taken into account in a repeated measures design.
There is a descriptive page here with some analysis options, although I disagree with the author that taking a simple change measure between baseline and test is the best method in this case.
Update: forgot about the lack of groups, thank you for pointing that out below, so yes the regression is the way to go. The baseline amplitude would be entered into the model as an independent variable. You will still need a model that allows a repeated measures design, otherwise it won't account for the within-subject variation as it will assume all observations are different subjects, and this assumption is not met with your design. With a binary measure for test amplitude (0 = no change compared to baseline, 1 = change compared to baseline), you will need to use logistic regression. It is important that you code your dependent variable as 0 or 1, not 1 or 2. If you instead decide that you wish to enter test amplitude as its value, instead of 0 or 1, then linear regression is one option. You'll need to make sure that the assumption of linearity is met. 
Update 2: Earlier I thought your model might have other factors such as age and sex included, so I was thinking that the control result would be a covariate. Because you have repeated measures, you can't do a simple linear regression because this model won't take account of the within-subject variance (it assumes each line in your dataset is independent, and this isn't the case). You have more than one independent variable - you have subject as a repeated measure, as well as the control amplitude. When you are using your software, you need to ensure that you include the repeated measure component. I can't advise how to do this specifically as I don't know what software you're using.
Update 3: I found this resource for repeated measures in SPSS. It looks like you want to find a regression procedure which has the REPEATED command as an option in the syntax. In R, have a look at the lme4 package, where you can fit subject as a random effect, which takes account of the repeated measures aspect, in particular look at the lmer command.
