Bit of a tester for the forum…but…

If I wanted to create a regression model of some data that has seasonal features (say, spikes in April and September) would it be OK to include 12 independent dummy variables switching between 0 and 1 to show up the seasonality and then look at the R2 for a given monthly variable to show importance?

Hmm.
I Suggest you drop a mail to RSwank, he might know. He's used to seasonally & otherwise variable stats.

PS. The new van is in progress, my bank account is nearly empty.

ta. and pics asap!

PS how you been? not seen your login around for a while…

Once it's ready I'll let you know.
I'm fine – and you? I lurked on MumsNet for a while, for the angst, but it became too depressing for words. Too much work didn't help, either

I lurked on MumsNet for a while

groomer 😉

Hmm, generally speaking you want to use the parsimonious model, that with least exogenous variables for a given adjusted R-2. There would be nothing wrong, however, with including 11 dummies (plus a constant) and strapping the model down by removing the statistically insignificant ones (from the t-test). But why? You state that your model has spikes in April and September… so I'd first try just two dummies in those two months. Are there troughs? Equally, you may wish to consider 3 dummies, plus one omitted case, for each of the seasons. It depends on the pattern of your data.

cheers jules – Ive had a go with 12 dummies and its not doing too bad but as you suggest I think trimming the variables down to just the expected peaks/troughs may allow the less clunky variables to work better – like Lag-12 and RPI. Thanks

42 ?