For a slight bit of insight, I'm (crudely) measuring my energy consumption (kWh and gas) per day. I have the thermostat set at a constant 19 degrees Celsius. However, I want to adjust the outcome for average outside temperature. How do I do that? I'm not modelling/predicting anything yet, so multivariable regression analysis does not seem applicable at this stage (but if that's the way to go, I'll do it).

My data is in has the following variables:

  1. Measurement date
  2. Electricity (kWh)
  3. Gas (m3)
  4. kWh/day
  5. m3/day
  6. Avg. outside temp in period since last measurement (degr. C)
  7. (TBD: electricity consumption/degree day)
  8. (TBD: gas consumption/degree day)

My idea was to calculate the consumption per degree day. Does this approach make sense statistically? Are there any alternatives?

Some background:

  • Gas is used for cooking, central heating, and heating water
  • Electricity is used for heating with an electric radiator
  • $\begingroup$ Degree day calculations just summarize temperature series in terms of sum (or integral) above or below some threshold. That does not provide any adjustment for anything else. So, it's hard to see what you are asking for here, as you say you are not asking about regression. $\endgroup$ – Nick Cox Nov 24 '15 at 10:26
  • $\begingroup$ Thanks for your feedback. I modified the question a bit: would the consumption per degree day make sense in this scenario? $\endgroup$ – Ben Nov 24 '15 at 10:40
  • $\begingroup$ So far as I can see you have only one predictor, the temperature, although there can be lag effects. So, some kind of regression or time series approach seems indicated, although your precise goals remain unclear to me. Quite how your electricity and gas are separate or linked only you can say. It seems quite likely that the response to temperature change would be different, e.g. depending on what was used for cooking, heating property, heating water, etc. $\endgroup$ – Nick Cox Nov 24 '15 at 10:46
  • $\begingroup$ Okay - basically electricity only affects heating with one electric radiator that is turned on occasionally. The rest is heated with gas. I guess though that at this stage this problem is not really suited for stackexchange. Thanks for your input anyway! $\endgroup$ – Ben Nov 24 '15 at 11:13
  • $\begingroup$ It could be made suitable. As your different versions of response variables imply, time scale modifies what you see. On the shortest time scales something is switched on or off and there's a stepped (and ramped) response. Over longer time scales (even daily) that is evidently averaged over, but temperature is variable too. Your systems will be responding to warmest and coldest temperatures as well as average temperature. $\endgroup$ – Nick Cox Nov 24 '15 at 11:38

You can convert your gas consumption in kWh so as to make it simple to compare it with your electricity use. Then you can run a correlation analysis with the temperature values you have to see the extent of the relationship. You'll need to make sure you have the same resolution of measurements for all your variables, hourly or daily.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.