Week 2: Data analytics thinking and cost estimation

Question 1

Cost estimation and it’s purposes

Answer 1

Cost estimation- well defined relationship between a cost object and its cost drivers for the main purpose of predicting costs.
Purposes:
1. Predict future costs using previously identified activity-based, volume-based, structural, executional drivers.
2. Identify the key cost drivers for a cost object.
3. Cost driver and cost-estimation relationship are useful in planning and decision making.

Question 2

6 steps in estimating cost function using quantitative analysis.

Answer 2

1. Choose the object to be predicted (DV) ex. production costs
2. Identify the cost driver (IV) ex. production quantity
3. Collect data on both variables
4. Graph the data
5. Estimate the cost function using one of 2 methods
   6.Evaluate the cost driver of the estimated cost function, accuracy (MAPE etc.)

Question 3

High-Low method:
Advantages and disadvantages

Answer 3

Advantages
- simple to calculate
- provides a first overview of the relationship between a cost and cost driver.
Disadvantages
- inefficient use of cost estimation (n=2) it ignores all other info.
-highly sensitive to outliers at the high/low points

Question 4

Regression analysis:
Advantages and disadvantages

Answer 4

Advantages
-uses info of all available observations
-can be adapted to non-linearity
-provides objective measure that can be judged on predefined criteria.
- minimised squared deviation
Disadvantages
-assumes that DV relates to explanatory variables,
correlation NOT EQUAL TO causation

Question 5

Evaluating regression analysis: R^2 (goodness of fit)

Answer 5

Indicates the explanatory power of the regression model, i.e. to what extent changes in the DV can be explained by changes in the IV(s)

Question 6

Evaluating regression analysis: SE (standard error)

Answer 6

Measures the dispersion of the actual observations around the regression line and thus gives a measure for accuracy of regression estimate.

Question 7

Evaluating regression analysis: p-value and t-value

Answer 7

t-value measures validity of independent variable in predicting dependent variable. Divide coefficient (slope) by standard error to calculate. Intuition: The higher, the better predictor

p-value is a translation of t-value into probability that observed relationship is due to chance

Question 7

Evaluating regression analysis: CI (confidence interval)

Answer 7

Measures accuracy of the prediction at any point for an IV. Unknown expected values are supposed to fall in this interval.

Question 8

Implementation problems: non-linearity. possible solutions.

Answer 8

High-low method cannot be adapted for non-linearity, but regression analysis can.
Issues:
1. trend and seasonality
* use price index (outside sources)
* trend variable
* replacement of original values of each values with first differences
2. Outliers
*dummy variable
3. Data shift
*dummy variable to indicate before and after shift

Question 9

Time-series regression

Answer 9

application of regression analysis to predict future amounts, using period data.

Question 10

Cross-sectional regression

Answer 10

estimates cost for a cost object based on info on other cost objectives and variables. Info taken from the same time period.