This problem also appears if I try the lag function in the lm function for a simpler case with less variables.
With "lm (EPEX~lag(EPEX,1), data = Koeffizienten)" I try to represent a autoregressive term in a linear regression, as you recommended above. To make it simpler I omitted all other terms, at the point we are at the moment we don´t need them.
I can´t find anything funny in the plot.
I repeat my objective again, because I´m not sure if it´s clear right now. I want to have a regression model, which calculates the powerprice by considering linear and autoregressive terms. Autoregressive means (for me), that the calculated price depends on the calculated price (from the regression model) from a privious time step.
It sounds logical to use the lag-function to achieve this objective, but it isn´t working as you see. Maybe it´s a kind of boundary value problem, because of the first hour where no privious hour is defined. But even than, I don´t have an idea how to solve it.
PS: Sry for the late response. I really appreciate your effort!