In ahardisty/dstrain:

library(ggplot2)
knitr::read_chunk('gas_eda.R')
knitr::read_chunk('gas_model.R')

Finding temperature trends for modeling on training set

Training set is usage and temperature data December 2013 for the W climate zone code.

## comparing temperature in winter months in 2013 and 2014

6am and 6pm are likely times when the majority of customers are home and will heat their homes in the winter.

We will use these two times of the day to test alternative modeling specifications.

Prediction power of Hourly vs. Daily Temperature

We used the training set of 2013 to compare the prediction power of hourly temperature (e.g., 6am and 6pm) to daily temperature for the prediction same day. On visual inspection the predictions and residuals from both models are identical. Differences between R2 values are negligible and as a result we will prefer the daily temperature for gas usage predictions since it is a simpler model.

ahardisty/dstrain documentation built on May 28, 2019, 4:42 p.m.