# Can I smooth my dependent data before using multiple linear regression?

#### megad

##### New Member
I'm working on multiple linear regression for energy analysis.. I have attached my data (graph - original & smoothing). Since the original data is highly fluctuated, so, I want to do smoothing and do regression after that.. I have 12 independent variables.. Is my method acceptable? I really need your advice.
Thank you

#### Attachments

• 45.2 KB Views: 4

#### hlsmith

##### Not a robit
Is this time series data?

Last edited:

#### megad

##### New Member
Is this time seriesdata?
yes..I have 14 independents variables..(surface temperature ).. my dependent variable is energy..

#### Attachments

• 38.8 KB Views: 2
• 60.1 KB Views: 2

#### hlsmith

##### Not a robit
This seems like very particular data. It is my understanding that you typically try to make a series stationary, fit a model, and estimates are kind of back transformed. Your variability does not seem stochastic, it seems like it is associated with real phenomena. Can you include indicator terms to address the spikes? I have not worked with data resembling yours before, but would wonder if you go to your field's literature, how do they model similar data?

#### megad

##### New Member
I’m working on the multilinear regression to predict the energy consumption in the office. My independents' variables are relative humidity, ambient temperature, and the surfaces temperature. I’m trying to create a model by using multilinear regression to predict the energy consumption for air conditioning system. Since I have more 14 potential variables, my initial assumption, I believe I can use multilinear regression.
The spikes in the data are not useful because the compressor needs more energy to start before it can run in normal mode. For the energy, the calculation is in Watt/hour (average of energy in 1 hour). However, my data is in every minute. What I have done so far is to use exponential smoothing to remove the spikes and allow regression analysis to predict the correct model, but I’m not sure whether the method is correct.