business analytics quiz 2

Standard deviation

• square root of variance

• square root of average squared deviations from the mean

how to increase r²

• Use more info

• Include multiple independent variables

• Still only ONE DEPENDENT VARIABLE

• So basically, multiple regression models

multiple regression model

time series

a data set where the data were obtained at regular time intervals

ex of time series data: monthly unemployment figures, hourly visits to a website, quarterly sales figures

— these kinds of techniques also known as TIME SERIES ANALYSIS

Casual or explanatory forecasting

set of quantitative techniques

these models usually rely on some kind of regression model to predict a variable of interest based on other variables that are assumed to cause or at least affect it

STATIONARY time series

time series that have a constant mean and a constant amount of random variation around that mean

Key takeaways: MULTIPLE REGRESSION

• r² almost always increases

• r² never decreases

• slopes change depending on model (independent variables are not truly independent, slopes depend on what else is in the model)

• interpret in context of full model (always say “in this model”)

when do you suspect multicollinearity ?

when slopes dont make sense

regular r² vs adjusted r²

adjusted r² is a skeptical judge —penalty for using too many variables

how to use adjusted r²

if adj r² inc → the new variable added real value

if adj r² dec → the new variable was just noise

if adj r² stays flat → the variable is redundant

how to determine number of dummy variables

= number of groups - 1

reference group

where both dummies = 0

curve fitting: log transformation

replace x with ln(x)

interaction term =

product of two variables

ex: gender dummy * years