Model dataset

Hi, i'm new and a bit of a novice.

Just wondering if there's some way to create a dataset based on a simple model - just a trend over time and constant - from which i can work out expected upper and lower quantiles for each period and thus see which observations in a real world dataset are abnormal given this simple model. This is so i can better see where a more complex model might be falling down in pushing up my r^2 and better specify my dataset.

This might be a very roundabout way of doing things but i tend to be pretty obtuse. Any more practical suggestions would be useful but i have around 20 variables and no brute force ways of finding the best specification/functional form - just feeling in the dark.

I've got givewin with the pcgive module, R and excel available to me...

thanks in advance :yup: