AARRGGHHH! I've had enough of this! There is nothing wrong with the protocol, there is nothing wrong with the measurement system (except it's older than my grandfather!) and there is nothing wrong with the method of analysing the data (because there is only one way to do it!) I've got months of work that looks ABSOLUTELY CR*P!!!! GRRRRRRRRRRRRRRRRRR!!!!!! Dammit, bloody GRRRRRRRRRRRR!

Calm. Must be calm. Anyone else want to join my in my holistic anti-"I hate this bloody PhD" therapy?

when you say its c**p? do you mean the data doesnt match the hypothesis? that it gives the opposite of what you would expect?
what is the data actually showing? maybe the data isn't crap. maybe it's telling you *something*

Know just what you mean.

You put the machine on. You go home and try to forget about it. You come in the next day, full of irrational hope, and - yes! - it's still c**p.

Ah Lara - I would so like to go with your idea! However, it's just c**p. I've fitted a Gaussian curve to test the data and about the only thing THAT line would be useful for would be as a kiddies water slide. However, in the interest of science, I will just make up some random new theory and muddy the waters of science for generations to come. Bu*$er it. Pantspantpants. Why oh why am I not allowed to unleash the awsome power of the English language manifesting as cussing on this website!

oh sorry to hear that. if i were in your shoes, i would feel exactly the same, very frustrated and wanting to break things.

but yah good idea, once you've let out all your frustrations, come up with an idea or a plan to figure out what you can do instead. good luck!

lol just noticed we can't say c**p - my post has been edited. wow the admin are fast!

Hypothesis- why are you fitting a normal (gaussian) curve? If your data is skew then it will never fit! If it is count data (0,1,2,3..) then have you tried a poisson curve? Or maybe you are trying to test the assumptions of your analysis in which case taking logs, or square rooting or recipricol sometimes makes the data more normal?

Given that I am working with a 250 x 150+ matrix per sample I'm using a Gaussian to plot the distribution of the data - make sure that the results are not horrendously out of sync with the mean. However, variability within the sample population, age of the machine, the fact I skipped breakfast this morning are all conspiring to ruin my data