Ah well, that's a bummer...
This is a plot of some preliminary data from a series of simulations I ran to test some analysis... the curve should show 1 direction (be monotonic)...
it's clearly not. **sigh**

This is science.

So what do a mean by 'simulations to test some analysis'?
This thread will be a describe how we use mathematics & computer coding to try and better understand neuroscience.

After reading lots of papers & speaking to experimentalists, we had an idea of what signals seems to effect the strength of certain connections in the brain. The experiments say: "if you make this cell active and this one quiet, the connection grows/shrinks" etc.

We write down some mathematical equations which (we hope) roughly capture this behaviour.
Often these are some form of ordinary/partial differential equations: describing how quantities change in time, depending on other quantities.

(this is an actual whiteboard of our musings)

'Analysis' then means we play with these equations, manipulating them into simplified forms to make some predictions about what we _should_ see when we write computer programs to solve them.
i.e. "if this quantity is very small, then the solutions should look like this..."

We do this so that, when we simulate the system, we have something to compare to. Otherwise, the simulation is just a bunch of numbers! How do you interpret it?
Once our analysis & simulations match up, we go back to the biology, and say 'ok, what parameter values should we use?'

This is called fitting the model to data.

But today, my simulations do _not_ agree with my analysis.
So I'll re-visit the maths, check I've done things right and try come up with some additional, testable predictions.

I'll then go back to my 'in silica' experiments: and expand the code to do more tests. Hopefully something, somewhere will give me a clue as to what's up...

How do you approach problem solving in your research/work environment?