Weighted Average Over Time for Data Projection (aka Weighted Moving Average) - Mathematically Correct Way to Determine Weights?

24 Views Asked by At

Consider the following data set:

Say graduation rates for a specific college's undergraduate program that are the following:

Year # of Graduates
Year 11 10,500
Year 12 10,750
Year 13 11,000
Year 14 11,750
Year 15 9,000
Year 16 11,500
Year 17 12,125
Projected # of Graduates
Year 18 ?
Year 19 ?
Year 20 ?
Year 21 ?
Year 22 ?

If I wanted to generate a data projection for the next five years using only these seven data points given; specifically, projecting Year 18 to Year 22 (five projected data points); what would be the mathematically correct way for choosing the weights to mathematically correctly generate these projected five data points?

Specifically, given the following mathematical formula for weighted average (${{x}_{w}}$) as the sum of the: products of weights ($w$) of values ($v$), divided by the sum of the weights, given as:

${{x}_{w}}={[(w_1)(v_1)+(w_2)(v_2)+(w_3)(v_3)+{\cdot}{\cdot}{\cdot}+(w_n)(v_n) ]/[(w_1)+(w_2)+(w_3)+{\cdot}{\cdot}{\cdot}+(w_n)]}$

Then how do you mathematically correctly determine the values of the weights?

Remark: I'm sure this has something to do with time scale duration between each of the data points in this set; which in this case are equal to one another at 1 year each (365 days); but I don't know how that yields the mathematically correct determination of the weights?

Any help on this would be appreciated, thanks in advance.

1

There are 1 best solutions below

0
On

There is no "mathematically correct" way. This is not a matter of mathematics. It is a matter of what model you believe best captures reality. The model you choose is based on your belief about how reality works -- not on mathematics.