In a classic DID model, the policy (or the treatment status) stay, once switched on, the same across the observation periods. That is, you have some units that are all untreated before the observation window. Then during the observation window, some received a treatment (the treatment status is 'on'), others stayed untreated (the treatment status is 'off').
I was reading an introduction of DID, on page 30 it says. There is a 'generalized DID model' that is more flexible. One of the flexibility is Switching (on/off) is allowed. However, it does not give any reference. Could anyone point me to a reference (prior study) paper, which used generalized DID to deal with a treatment that has switching throughout the periods? Thank you!