Time Series Data and Forecasting Flashcards
what is time series data
numerical data in regular intervals
eg annually, quarterly, monthly, weekly, daily etc
which axis is always the time period
x
which axis is always the variable of interest
Y
What is a trend
a gradual upward or downward movement (persistant)
can trends be non linear
yes so long as they are overall moving in the same direction in total
what are seasonal movements
short term waves often monthly or quaterly eg sale of school clothes
what are cyclical movements
long term waves that regularly occur and vary in length eg house prices
2 ways of smoothing data
moving averages, exponential smoothening
what do the results of moving averages depend on
L (length of time)
if there are bigger windows of time for L is there more or less errors
more
why might you use moving average
easier to see underlying trends
what does MAD stand for
Mean absolute difference
are we looking for a high or low MAD
as low as possible
what is exponential smoothing
weighted moving average
what two numbers do weights lie between
1 and 0
in exponential moving what is the first element
the result of the first one as there is no prior info
how weighted average is calculated for each element after the first one
weightactual + (1-weight)actual of previous
common pitfalls in time series analysis
assuming current trend will hold
not considering changing habits, technologies, business experience
what is time series analysis
using data in the past to predict the future
if using L=3 , how would you get moving average for the year 2002
average(2001-2003)
what is the moving average prediction for the year 2002
2001 smoothened version
