residuals to an expected distribution

I've two variables which are linearly related. I made linear regression, getting the estimated slope and intercept. I want to test the null hypothesis that the estimated regression line is different from a theoretical line. I've already used t-student tests for the slope and intercept, but, given the low size of the sample (11 data points were available for the linear fit), I'd like to use also non-parametric tests. I'm doing these calculations with the residuals of experimental data from theoretical values; I've used sign test, Wilcoxon-signed-ranks test, and Wald-Wolfowitz test (the latter according to the size of the starting variables). I'd like to use also Fisher exact test, with 2x2 matrices; the point is: while for the columns I can share data depending on the signs of their residuals, is it correct to share rows according to the magnitude of one among the two starting variables?
I hope that the question was clear!!:p