The household survey included 40 questions on the attitudes of
occupants towards energy consumption. Using Factor Analysis as
was explained in previous sections, and informed by behavioral
sciences research, we formed 22 major factors that collectively
explain more than 80% of the information included in the original
40 questions; i.e., the threshold value in the scree test for choosing
the number of factors was explaining 80% of the variance. The 22
variables explain the attitudes of households in three groups: (1)
Energy Efficiency Actions, (2) Information Seeking Behaviors, and
(3) Home Improvements Behaviors. After Factor Analysis, adding a
number of transformations of the original variables, and removing
some variables, the total number of variables was reduced from 114
to 97. We fit separate models to daily maximum, minimum,
maximum minus minimum, and average consumption, both for
summer and winter (for the period when the data were available),
and ranked the variables by their importance through a forward
stepwise model selection procedure (Fig. 1).