Fixed Effects model with seasonality dummies

[Nick Simmons]

Hi all,

As a bit of background, an installation, i, produces heat, and has a heat load factor (a proportion of heat generated of its maximum capacity). This fluctuates throughout the year due to it being warmer in the summer (less heat is required) and colder in the winter. What I want to do is calculate a load factor for each i that has seasonality removed from this, in order to project this forwards to estimate heat (it is then adjusted by seasonality). My approach is:

• Calculate monthly load factors for each installation (i and t)
• Estimate a FE panel regression (for a particular tech that we assume has the same seasonality assumptions):y_it= B_0+(sum j=1 to11)(B_j*D_j)+a_i+u_it
• The aim is to get the fixed effect for each installation in order to use it to project forwards (when projecting forwards, heat (calculated using the load factor) will be adjusted by seasonality).

My main question is how do I calculate what the heat load factor is, for each i? My confusion arises because when I remove a different dummy variable, (a_i + u_it) is fixed, but the constant is not. Do I need to calculate a weighted average to get this?

Any help would be great.

[Nick Simmons]

To follow up on this, calculating the 'average' load factor for the tech is straightforward - multiplying the (dummy coefficient by the no. of days in each month)/365 + the constant gives the overall average. Calculating seasonality assumptions is also easy as it is the constant + the monthly coefficient, and then normalising.

My question is how do I calculate the load factor for each installation i? It is the overall average + (fe+ u_it)? Or does a second regression need to be estimated, fixing the seasonality assumptions, and then the first installation becomes 'the constant'?

[Phil Bromiley]

You'll increase your chances of a useful answer if you follow the FAQ on asking questions - provide Stata code using code delimiters, Stata output, and sample data using dataex. You should also strip out any unnecessary code to make it easier for us to read.

You question seems to be more about the substance of the issue than Stata. We are not expert in the substance - ask experts in your topic. We can help you implement either approach, but which is right is not a Stata issue per se.