Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
34 Cards in this Set
- Front
- Back
What are the two general types of forecast techniques?
|
Qualitative and Quantitative
|
|
What are the two sub-categories for each broad method?
|
Qualitative: Judgmental and Market Research
Quantitative: Time Series and Causal |
|
Qualitative Forecasting Methods
|
- involve greater subjectivity
- used when historical data are lacking -Judgmental and Market research methods |
|
Judgmental methods
|
- Qualitative
- based on experience and informed opinion -jury of executive opinion(ppl who constantly have access to external and internal info) - sales force estimation (have bias, req to come to consensus) - Delphi (anonymous experts) |
|
Delphi-
|
anonymous experts; independently give feedback on future. Summarize opinions of random sample
|
|
Market research methods
|
- Qualitative
- based on systematic data collection and statistical analysis - focus groups (selected carefully, representative of demographics) - consumer surveys (rely on large #s of people) |
|
Quantitative Forecasting Methods
|
- based on mathematical techniques
- requires historical data - more objective approach Time Series methods and Causal modeling method |
|
Time Series Forecasting Method
|
- Quantitative
- forecasts are based solely on past (historical) demand - assume future demand is reflected in past demand patterns - short term |
|
Causal Modeling Forecasting Method
|
- Quantitative
- demand is influenced by one or more independent variables - best for underlying cause - long term |
|
Advantages of Causal Modeling compared to Time Series Modeling:
|
- typically better for long-range forecasting
- typically better for predicting shifts (or turning points) in data patterns - can generate better understanding of the mechanisms influencing sales/ demand |
|
Disadvantages of Causal modeling compared to time series modeling:
|
- Causal requires more data
- Causal requires special expertise that many managers do not have |
|
5 Step Causal Modeling Procedure using Regression analysis:
|
1. identify variables that you "think" might explain what you are trying to forecast - initial (or theoretical) regression model
2. collect the data (observations) 3. initial screening of the data 4. build regression model and test - check that data meet necessary conditions 5. use model to forecast |
|
in the initial screening of the data, step 3, what can result in removal of certain problematic data points?
|
- errors in recording the value
- observation is not part of the population of interest Case 1: error in recording or reporting data value Case 2: data representing extreme or rare conditions such as -major disruption in supply chain -rapid change in market conditions problematic data points- should be removed prior to model development |
|
what is defined as "extreme observations" ?
|
an extreme observation, or OUTLIER, is identified as being well separated from the remainder of the data
-an observation may be extreme, or outlying, with respect to its dependent variable, it independent variable, or both - outliers warrant additional scrutiny |
|
how do you go about detecting observations?
aka Outlier Analysis: |
-use standardized residuals (errors of the regression equation)
- values greater than +-4 standard deviations from the mean indicate a potential outlier |
|
what are the four assumptions of regression?
(necessary conditions for regression analysis) |
a) linear relationship exists between the independent variables and dependent variables
b) independent variables are NOT highly correlated; correlation table IrI > 0.9 c) residuals (or errors) exhibit a constant variance; aka homoscedasticity (we want) d) independent residuals (good) if violated: (bad) 1) autocorrelation 2) serial correlation |
|
what two graphs check a linear relationship?
|
scatter plots and residual plots
|
|
scatter plots
|
-plots with dep var on the y-axis and each ind var on the x-axis
- a plot should suggest a straight line relationship (with or without a trend) |
|
residual plots
|
- plots with residuals on the y-axis against each ind variable and the predicted values on the x-axis
- a plot SHOULD exhibit a RANDOM pattern |
|
independent residuals
|
- the residuals must be independent of each other
- plot should NOT exhibit a trend or any special pattern - residuals should be distributed randomly between pos and negative pattern |
|
normally distributed residuals
|
- residuals must be normally distributed
- histogram should exhibit a bell shaped curve about zero, using 8 to 10 bin categories |
|
Naive Method
|
the naive forecast for the next period equals the demand for the current period.
a.k.a. what happens today will happen tomorrow |
|
Simple Moving Average (SMA)
|
(Horizontal; no trend)
- SMA places the same weight (or emphasis) on each time period - works well when demand is fairly STABLE over time - does NOT do a good job of forecasting when a TREND is present - the forecast LAGS THE ACTUAL DEMAND because of averaging effect - DECREASING the number of periods in a forecast, creates a MORE RESPONSIVE forecast |
|
Weighted Moving Average (WMA)
|
(horizontal; no trend)
- weighted moving average allows DIFFERENT emphasis to be placed on different time periods - works well when the demand is fairly stable over time - does NOT do a good job of forecasting when trend is present - the forecast lags the actual demand bc of averaging effect - weights tend to be based on the forecaster's experience - DECREASING the number of periods in forecasting and/or INCREASING the size of the weights for more recent demand creates a MORE responsive forecast |
|
what is forecasting error?
|
Et= (At-Ft) Forecast Error is the difference between actual demand and the forecast
- used to evaluate the accuracy of a forecasting model - for all measures of forecast accuracy, the CLOSER the measure is to ZERO the BETTER the forecast |
|
ultimate goal of any forecasting endeavor is to have an
|
accurate and unbiased forecast
|
|
Costs associated with prediction error inclue:
|
cost of lost sales, safety stock, unsatisfied customers, and loss of good will
|
|
Absolute measures of forecast include:
|
RSFE, MFE, MAD, MSE
|
|
Relative measures of forecast include:
|
MAPE, tracking signal
|
|
RSFE and MFE
Running Sum of Forecast Errors and Mean Forecast Error |
POSITIVE= underestimated demand; stockouts
NEGATIVE= overestimated demand; higher inventory costs - a zero RSFE and MFE indicates that the forecast is UNBIASED, but NOT necessarily accurate |
|
bias
|
represents the tendency of a forecast to be consistently higher or lower than the actual demand
|
|
MAD
|
Mean Absolute Deviation
- measures average magnitude or size of the forecast errors without regard to the direction of error - simple way to compare forecasts - greater than zero, overestimates or underestimates - a ZERO MAD= PERFECT forecast of demand, overlap |
|
MSE
|
mean squared error
- sensitive to large errors |
|
tracking signal
|
- used to monitor the performance of a forecast overtime
-tracks forecast bias relative to the average magnitude of the forecast error |