quantitive forecasting techniques Flashcards
what is the high low method
technique used to predict an outcome based on another β> predicting a dependant variable
when can we use the HL method to estimate the relationship between two variables
when they are quantifiable
what would a result of a HL calculation look like
y = a + bx
y= the value for the dependent variable
x = the value for the independent variable
a = that part of y which does not depend on x
b= how much y changes if x changes
two advantages of regression analysis
- It uses all the paired data to arrive at a definitive line of best fit.
- We can test the reliability of the analysis for forecasting by estimating
the degree of correlation there is between the dependent and independent variables.
formulae for regression line of best fit
b = ( nβπ₯y β βπ₯βy) /nβπ₯2 β (βx)2
a = βy/π - π (βx /π)
n = the number of pairs of data
what are the limitations of forecasts based on regression
- Not all relationships are linear
- Focus on two variables makes it more likely that relevant factors may
be ignored - Care should be taken outside of the relevant range β interpolation is
usually more reliable than extrapolation - The line of best fit may not have a high correlation
what is extrapolation vs interpolation
interpolation refers to determining something whilst the value is in range whereas extrapolation the value lies outside or range
what is correlation
Two variables are said to be correlated if a change in the value of one variable is accompanied by a change in the value of the other.
what is the correlation coefficient
- The correlation coefficient (r) measures the extent of the linear correlation between two variables.
- The correlation coefficient value will fall within the range of +1 to -1.
what does a correlation coefficient close to +1 represent
strong positive correlation
what does a correlation coefficient close to zero represent
whether positive or negative the weaker the correlation.
explain cause and effect
Correlation describes how one variable moves alongside another. It does not prove that the move in one variable causes a move in the other.
what is rank correlation coefficient
measures the correlation between two sets of ranking
when can rank correlation be more useful
when it is important to observe relative values of what is being measured rather than value themselves
spearmanβs rank correlation coefficient
p = 1 - (6βπ^2)/ (π(π^2 β 1))
d = difference between two ranks for an item
n = number of items ranked
explain learning effect theory
workforce gains experience in a task so will come to perform it quicker, labour costs reduced over time - not expected to be indefinite
conditions necessary for learning effect
- significant manual element
- repetitive
- early stage of production
- consistency in the workforce
- no extensive breaks in production
- motivation of the workers
definition of learning effect
every time the cumulative total output of a product doubles, the cumulative average time taken to make all the units to date fellas to a proportion of what it was beforehand
definition of learning rate
the proportion to which the cumulative average time per unit falls
learning curve formula
y = ax^b
a - time taken to produce the first unit
x - cumulative number of units
b - the index of learning (logLR/log2)
LR = the learning rate (as a decimal
what happens when steady sate is reached for time per unit
the time per unit or labour cost per unit becomes a standard cost that can be used for ongoing budgets
when does cessation or learning effect arise
usually when one of the assumptions underpinning the theory is no longer applying
what is a time series
a sequence of numbers, values, measurements recorded against a timeline
wha is benefit of time series analysis
allows observations to be made and conclusions to be drawn about how a variable behaves over time.
what are the types of variations which make values more away from trends
seasonal variations
cyclical variations
random variations
what is seasonal variations
short term fluctuations in values due to where we are in the planning cycle
what is cyclical variation
recurring patterns over a longer period of time
what are random variations
irregular or unpredictable variations
time series =
tren (Td) x seasonal variations (SV) x other variations (CV+RV)
was to calculate trend
- regression analysis
- moving averages
how can you estimate some SVs
have to assume that any difference between time series and trend is seasonal
limitations of time series analysis
- not all changes relate to time, seasons and cycles
- further ahead we try to forecast, the less reliable the forecast
- historical patterns cannot always be assumes
