They are upgrading the #software running a #radar I work with. One of the tests is obviously how good the data looks. The new software has higher standard deviations than the old software (#statistics)

And that's it. They are trying to debug it based on just the std dev.

I looked at one of the tracks and there was clearly a time dependent trend.

I keep telling them and telling them that just thinking in terms of "standard deviation" is making them assume gaussian #noise sources.

But if you look at the #data as a *time series*, you will start thinking about the processing and what would be happening at different points in a track and how it might go wrong starting at certain points.

It's a purely software #engineer team--they know about radar (more than I do!), but aren't #system engineers.

It's been going on long enough that I want to say something to someone higher up. But I don't want to adopt another project....