Sie sind auf Seite 1von 24

BUSINESS REPORT EXCLUSIVELY FOR

Mohammed H. Khan 2nd Feb. 11

Student Number: 08178887

Analysis of Current Issues in Finance Word Limit


1500 (excluding references)

Contents Page:
Section Page No.
Executive Summary 3

Introduction 4

Section 1 – Decision Analysis 4–6

Section 2a – Time series forecasting 6–7

Section 2b – Linear regression analysis 7

Conclusion 7

Appendices

Appendix A – Payoff table 8

Appendix B – Influence Diagrams 8

Appendix C - Minimax Regret approach 9 - 10

Appendix D - Expected value approach 10

Appendix E - Expected value with 11


perfect information

Appendix F - Sensitivity Analysis 11 – 12

Appendix G - Moving Average Example 13 - 14


Executive Summary

This report provides an analysis on how decision and time series analysis can be
used in financial situations. Decision analysis (DA) provides a systematic
approach to obtain an optimal decision strategy when managers are flummoxed
with decisions. Time series forecasting involves using historical data to predict
future outcomes.

DA dissects complex decisions into manageable components, a very effective


way to deduce an optimal decision. However implementing DA requires specialist
knowledge both time consuming and expensive.
Devising probabilities for new projects are likely to be subjective and erroneous
especially when historical data is unavailable, prescribing incorrect probabilities
is problematic however DA can be used to determine in advanced likely effects
of errors in probabilities associated to the optimal decision via sensitivity
analysis.

Time series is a set of observations on a variable measured at successive points


in time. Three smoothing methods are employed to eradicate random
fluctuations in a time series which exhibits no trend. The moving average
consists of computing an average of past values and then using this average as
the forecast for the next period. The weighted moving average allows for
different weights on data and exponential smoothing involves only one weight
the weight of for the most recent observation. Weights for other data values are
automatically computed and get smaller as observations move further into the
past. However, moving averages cannot pick up trends very well – they will
always stay within past levels and not predict a change to a higher or lower level
(Time lag). When time series data is not available regression can be used for
forecasting. Regression involves deducing an equation which shows how
variables are related.
Time series is a predictive method based on extrapolations, which, for the near
future the data set maybe accurate, but the further away from the observed
data the more error is introduced and the less accurate the model becomes
making it ineffective for long term investors.
Introduction

The raison d’être of the report to critically analyse how Odey Asset Management
can apply decision analysis and time series forecasting to their business.
The report is split into two sections section 1 explores decision analysis providing
analysis on how problems are structured using payoff tables, tree diagrams and
influence diagrams.
Section two is split into two sub-sections where section 2a discusses time series
analysis and section 2b explains how regression is used to devise a forecasting
equation.

Section 1 – Decision Analysis


Managers are flummoxed with many decisions, thus require a pragmatic
approach to identify optimal decision strategies, this is where decision analysis is
used.

Investment Successful (£m) Unsuccessful (£m)


(s1) (- indicates loss)
(s2)
Equities (d1) 8 7
Commodities (d2) 14 5
Emerging Markets (EM) (d3) 20 -9

Say Odey Asset Management dilemma is to decide which investment to select.

d denotes decision
s denotes situation

DA offers three ways to structure problems.

Payoff Tables: used to select the best alternative given a range of outcomes. A
payoff table shows expected payoffs for each possible outcome. (Ozcan: 2009) (See
appendix A)
The payoff table shows profit which is denoted millions. If the EM is chosen and is
successful, a profit of £20 million will be earned where as if it is unsuccessful a loss of
£-9m .
S1 S2
d1 8 7
d2 14 5
d3 20 -9

Influence diagrams: are visual representations of the model. They allow models to
be built in
parts and the effects of various parts to be seen without “getting immersed in the
details of the model”. (Crundwell : 2008) (See appendix B)

States of nature
Outcome
Successful
Unsuccessful

Profit
Investm
Consequences
Decision Alternative Profit
Equity
Commodities
EM

Decision trees are graphical representations of decision problems that show the
chronological nature of the decision making process.
Experts in problem solving suggest complex problems are best solved by
initially dissecting the problem into smaller sub problems “(divide and conquer)”
illustrating the stages of the decision process which makes it easier for decision
makers to justify their decisions. (Anderson et.al: 2008, Poulton: 1994)

Successful 8
Equity 2

Unsucessful
7

Successful
1

Commodies
3
1
Unsucessful
5

Successful
2
Emerging Markets
4

Unsucessful
-

However, in practise multiple complex decision alternatives are difficult to


reduce into a list of outcomes. (Friedland: 2000)

DA avoids optimistic managers getting carried away with most likely outcomes
and encourages them to recognise the range of possibilities (Pike et.al:2006).

DA puts forward 3 approaches to decision making without probabilities.

Optimistic approach – For a maximisation problem, it chooses the decision


alternative corresponding to the highest payoff. (Selects EM)

Conservative approach: recommends a decision alternative which provides the


best of the worse payoffs. (Selects equity)

Minimax Regret approach: minimizes the worst-case regret .The regret is the
payoff which a decision maker misses by making the wrong decision (opportunity
loss) (Investopedia: 2010). It puts forward the decision alternative which
minimises the maximum regret. (Select commodities). (See appendix C).

When probabilities are provided for the outcomes, expected value approach is
used to identify the optimal decision. (See Appendix D) . Associating probabilities
to a new project can be difficult when there is no previous data available.
Probabilities are likely to be subjective inspired by guesswork which can lead to
incorrect decisions. (Aven et.al : 2010) .

DA can provide the potential value of research which say obtained perfect
probabilities of an outcome occurring and thus discloses the extra value gained
from an investment as a result. (See appendix E)

Sensitivity analysis can determine how changes in probabilities for the different
outcomes affects the optimal decision, providing hindsight to the analyst to
acknowledge all outcomes whilst informing them about the room they have for
judgemental error and to decide whether they are prepared to accept risks. (Pike
et.al : 2006) (See appendix F)

Section 2a – Time Series Forecasting


A time series (TS) is a set of observations of a variable e.g sales measured at
successive points in time or over successive periods of time. (Hennessy :1989) .
Forecasting is the process of analyzing current and historical data to determine
future trends. (Investopedia: 2011)

Smoothing techniques are used to reduce irregularities (random fluctuations) in


time series data and provide a clearer view of the true underlying behaviour of
the series.

Smoothing methods are only appropriate for a stable time series, one which does
not exhibit a significant trend, cyclical or seasonal effects (Statistical glossary:
2011). Smoothing methods are easy to use and provide a high level of accuracy
for short range forecasts.

Moving average (MA)

This method uses the average of the most recent data values in the time series
as the forecast for the next time period. When a new observation becomes
available, it replaces the old observation and a new average is derived. However,
MAN ignores the effect of any data older then ‘N’ periods. (Anderson et al: 2001)
(See appendix G)

Weighed moving averages

A moving average that is weighted so that recent values are more heavily
weighted than values further in the past, when the time series is volatile
selecting equal weights to each data value maybe best (See appendix H)

Increasing the size of n1 smoothes out fluctuations better, but makes this
method less sensitive to real changes in the data (Anbuvelan:2007)

Exponential smoothing

A weighted average in which more recent data is given greater weight than older
data, weights for older data values automatically get smaller and smaller as the
observations move further into the past. Exponential smoothing does not require
all past data to be stored in order to compute future forecasts reducing errors
and requires minimal data which is excellent when forecasts are required for
myriad of items. (See appendix I)

Problem with the smoothing methods is they cannot pick up trends well – they
will always stay within past levels and cannot predict a change to a higher or
lower level (Time lag)

1 n is number of periods
The accuracy of the forecasts derived by the smoothing methods can be
measured using the mean squared error, which in theory allows the analyst to
perfect forecasting method (See appendix G)

Section 2b – Linear regression


Linear regression is a statistical tool used to develop an equation showing how
variables are related (Statistical glossary: 2011). Regression analysis is beneficial
as it can also be used when time series data is not available or when a time
series consist of random fluctuations around a long term trend line.

If Odeys firm received funding which was inversely related to interest rate, Odey
could forecast how much investment his firm would receive at specific rates of
interest. (See appendix J & K – for calculations)

Analysts should be wary about decisions based on forecasts outside the range of
known values (extrapolation). Although, the estimated regression equation
represents the best possible linear relationship for the data set within the limits
of the data there is no guarantee the relationship will be valid outside the data
set (Hennessy: 1989).

Conclusion
DA is useful when faced with many decisions, as it dissects problems into
manageable tasks.
When using the expected value to determine optimal decision, caution should be
exercised when using subjective probabilities. Sensitivity analysis should be used
to ensure the stability of the value of the optimal decision given a change in the
inputs.

With the time series methods, trends cannot not be predicted but confirmed. If
time series data is not available regression analysis is used to identify a
relationship between variables.

TS assumes the future looks like the past and is a predictive method based on
extrapolations which for the near future the data set maybe accurate, but the
further away from the observed data the more error is introduced and the less
accurate the model becomes.
Time series analysis may not be beneficial for financial firms who require
forecasts for long time periods as in the case for Odey asset management; a
long term investor.
Appendices
Appendix A – Payoff table

The payoff table shows the expected return for each outcome under a specific
condition, referred to as states of nature.

The table is constructed using the alternative outcomes represented as An ,


with the conditions represented as Sn and outcomes represented as On .

(Source: Ozcan:

Appendix B – Influence Diagrams

An influence diagram uses shapes known as nodes and arrows known as arcs
which allows the diagram to function as a representation of the system.

Nodes represent decisions (chance events) and consequences. Lines


connecting nodes known as arcs represent influence between variables.
The direction the arc is important, as it specifies that the value of the node at
its head (arrow end) depends directly on the value of the node at its tail
(Decision Craft: 2010).

Source: Crundwell :
2008
The shape of the nodes represents its purpose.

Rectangle or square nodes represent decision nodes. A decision is a variable


that the organisation can change independently.

Circles or ovals represent chance nodes. A chance variable is uncertain;


something the company will not be able to predict the outcome of.
Appendix C – Minimax Regret approach

The Minimax Regret Approach puts forward decisions after evaluating the regret
linked with each alternative. The regret is the payoff which a decision maker
misses by making the wrong decision The regret Rij associated with decision di
and state of nature Sj is worked out by:

Rij = Vj * −Vij

Where

Rij: Regret associated with decision di and state of nature Sj


Vj*: The maximum payoff under state of nature Sj
Vij: Payoff associated with decision di and state of nature Sj.

When the regret matrix is deduced, the next step is to determine the maximum
regret associated with each decision di. The optimal decision d* is the one which
generates the minimum regret among the maximum regrets evaluated for each
decision.

R* = Min{Max Rij }
di Sj

This approach is suitable for a decision maker who generally has regrets after

Regret table for Odey Investment


Decision alternatives States of nature
Successful Unsuccessful
Equity 20-8 = 12 0
Commodities 20-14 = 6 7-5 = 2
Emerging markets 20-20 = 0 7 - -9= 16

making a decision which turns out not to be best possible one. (Lebcir: 2010)

The minimax approach can be used to work out what investment Odey should
select.
Supposing Odey decides to invest in equity and it is successful thus making a
profit of £8m
However, if strong
Maximum regret for each Odey Investment demand was
alternative decision unavoidable with any of
Decision alternatives Maximum Regret their products, Odey
should have invested in
Equity 12 EM as it yields the
Commodities 6 greatest profit of £20 m.
Emerging markets 16 The difference between
the best decision payoff
(£20m) and the decision to invest in equity (£8 m) is the decision regret. So, the
regret or the opportunity loss is £20 million – £8 million = £12 million.

To workout, the regret or opportunity loss associated with each combination of


decision alternatives and state of nature, we subtract each entry in a column
from the largest entry in the column . The table below shows the regret for
Odeys dilemma.

Minimum of the
maximum
regret

To derive the minimax decision, is to list the maximum regret for each decision
alternative. According to the Minimax approach, Odey should invest in
commodities.

Appendix D – Expected value approach

In this approach, the expected value of each decision is worked out and the
decision alternative which generates the highest expected value is selected. The
expected value for each decision di is calculated as follows:

N = the number of states of nature

P(SJ) probability for an outcome occurring

As one outcome can only occur, probability must satisfy two conditions:
– P(sj) ≥ 0 for all outcomes
– Probabilities of all outcomes must sum up to 1.

Thus:

EV FORMULA = j=1NP(Sj)Vij

If Odey is optimistic about the potential of the all three funds performing so as to
give them a subjective probability of 0.8 of being successful and thus 0.2 for
being unsuccessful.

EV of each outcome is =

EV(d1) = 0.8(8) + 0.2(2) = 7.8


EV(d2) = 0.8(14) + 0.2(5) = 12.2
EV(d3) = 0.8(20) + 0.2(-9) = 14.2

EV suggests Emerging Markets should be selected.

Appendix E –Expected value with perfect information

If Odey was certain that all funds were going to be either successful or
unsuccessful, d3 would be chosen as it yields the highest profit when successful
and d1 would be chosen as yields the highest minimum profit when
unsuccessful. Therefore the expected value with perfect information (EVPI):

0.8(20) + 0.2(7) = 17.4 = EVPI

Without the perfect information, fund 3 was selected with an EV of £14.2m.

£17.4m – £14.2m = £3.2m which is the additional expected value obtained if


perfect information is available about likelihood of outcomes occurring.

Although, activities associated to gathering perfect information such as market


research may not yield perfect information, the information gathered maybe a
worth a sizeable portion of the cost derived by the EVPI. (Anderson et.al:2008)
Appendix F – Sensitivity Analysis

The expected values, which constitute the criteria for decision making under
uncertainty depend on the estimates of the probabilities associated with the
states of nature. Therefore, if these estimates change, the expected values will
also change, which may lead to a change in the decision. The process of
evaluating the effect of changes in the state of nature probabilities on the
decision process is called sensitivity analysis.

Sensitivity analysis can be employed by firms to see how changes in probability


of an outcome can affect the decision. For example, say the probability of the 3
funds being successful was changed to 0.2 and the probability of them being
unsuccessful was 0.8.

New EV for three outcomes are:

EV(d1) = 0.2(8) + 0.8(7) = 7.2


EV(d2) = 0.2(14) + 0.8(5) = 6.8
EV(d3) = 0.2(20) + 0.8(-9) = -3.2

Recommended decision alternative is to invest in equity.

This means when there is a higher chance that all investments are going to be
successful then Odey would select EM. When there is a higher chance of the
investment being unsuccessful then Odey would invest in equity.

A graphical method can be used to determine how changes in probabilities of an


outcome affect the investment decision. .

For example, let p = probability of state of nature S1 thus P(s1) = p . as there


are two states of nature, the state of being successful and unsuccessful thus:

P (s2) = 1 –P(s1) = 1- p

If =j=1NP(Sj)Vij

Now we develop 3 equations which show EV of three decision alternatives as a


function of the probability of S1
To make P a function for each decision alternative:

EV(d1) = P(s1)(8) + P(s2)(7) (1)


= p(8) + (1-9)(7)
= 8p + 7 -7p = p+ 7

EV(d2) =9p+5 (2)


(3)
EV(d3) = 29p – 9
Creating a graph with P on X axis and EV on Y axis, as above equation is linear
graph of each equation will be a straight line.
So, if P = 0 in equation 1 then EV(d1) = 7. If P=1 then EV(d1) = 8. Connecting the
two (0,7) (1,8) points on the graph provides the line labelled EV(d1). This process
will be repeated for EV(d2) and EV(d3) .

The graph shows how the optimum decision alternative changes as probability
for the funds being successful changes.
The value of P for which EVd1 = EVd2 is the value of p when EV(d1) and EV(d2)
intersect.
P=

P + 7= 9p + 5
8p = 2
P = 28 = 0.25

Value of p when EV (d2) and EV (d3) intersect = 0.7

Thus:
d1 provides largest EV when p ≤ 0.25
d2 provides largest EV when 0.25 ≤ p ≤ 0.70
d3 provides largest EV when p ≥ 0.70

Probability for S2 =
– p is the probability for s1
– so (1-p)

Appendix G – Moving Average Example

MA uses the average of the most recent data values in the time series as
the forecast for the next time period.

Moving average = most recent n data valuesn


Week Barclays share price The share price displays random
(p) variability but shows time series to be
stable over time.

To use the moving average to forecast


1 17
Barclays share price, we first select the
2 21 number of data values to be included in
the moving average.
3 19
To compute say a 3 week moving
4 23
average:
5 18
Moving average (weeks 1-3) =
6 16 17+21+19 3 = 19

7 20 19 is the moving average forecast for


week 4 . However, the actual value for
8 18
week 4 is 23, so the forecast error is 23-
9 22 19 = 4.

10 20 To calculate the second 3 week moving


average is :
11 15
Moving average (weeks 2-4) = 21 + 19
+ 233 = 21

Forecast for week 4 is 21, error associated with this forecast is 18 – 21 = -3

A forecast can even be made outside the data range for example, the forecast
for week 13 would be:

Moving average (weeks 10 – 12) = 20 + 15 + 223=19

Forecast for week 13 is equal to the average of the share prices for weeks 10, 11
and 12 . The forecast for week 13 is thus 19 pence.

Summary of three week moving average calculations

Week Time series value Moving Average Forecast Error Squared Forecast
error
Forecast

1 17

2 21

3 19 19

4 23 21 4 16

5 18 20 -3 9
6 16 19 -4 16

7 20 18 1 1

8 18 18 0 0

9 22 20 4 16

10 20 20 0 0

11 15 19 -5 25

Total -3 83

It is important to consider accuracy of the forecast when selecting a forecasting


model. Ideally, the forecast error should be low, the last two columns contain the
forecast errors and forecast errors squared, which are used to measure the
forecast accuracy.

Mean squared error (MSE) is used to measure the accuracy of the forecast and is
calculated as:

Average of the sum of squared deviation: 838 = 9.2

Using different lengths for moving averages will affect the accuracy of the
forecast. Thus, the analyst should use trial and error to identify which length
minimises MSE so as to obtain reliable forecasts. (Anderson et al.: 2008)

Appendix H - Weighed moving average (WMA)

Weighed moving average is a smoothing method that uses a weighted average


of the most recent n data values as the forecast.

Calculating a 3 weighed 3 week WMA using Barclays share price with most
recent observation receiving a weight three times as great as that given to the
oldest, and the next oldest receiving a weight twice as great as the oldest. So for
week 4 the calculation would be:

Weighted moving average forecast for week 4 = 36(19) + 2621+ 1617= 19.33

Note: Weights have to sum up to 1.

If the analyst believed that the recent past is a better predictor of the future than
the distant past, larger weights can be given to the more recent observation.

To determine which combination of data values and weights provide the most
accurate forecasts the MSE can be used. The combination of data values and
weights which derive the lowest MSE for historical time series should be used to
forecast future values in the time series. (Anderson et al.: 2008)
Appendix I - Exponential smoothing

Exponential smoothing: A statistical technique that calculates a moving average


where the most recent data is given a different weight to earlier data (ESOMAR:
2011).

Exponential smoothing equation

Ft+1 = aYt + (1 – a)F1

Ft+1 = Forecast of the time series for the period t+1


Yt = actual value of the time series in period t
F1 = forecast of the time series for period t
a = smoothing constant (between 0 & 1) - For example if µ = 0.2 then this
indicates that 20% of the weight in generating forecasts is assigned to the most
recent observation and the remaining 80% to previous observations.

Week Barclays share


Although, exponential smoothing provides a price (p)
forecast that is the weighed average of all the past
observations, all past data will not be needed to
calculate forecasts for the next period. 1 17

With a smoothing constant only two pieces of data is 2 21


required for forecasting. 3 19

4 23

5 18

6 16

7 20

8 18

9 22

10 20

11 15
Week (t) Time series value Exponential Forecast error
(Yt) smoothing
forecast (Ft) (Yt – Ft)

1 17

2 21 17 4

3 19 17.80 1.2

4 23 18.04 4.96

5 18 19.03 -1.03

6 16 18.83 -2.83

7 20 18.26 1.74

8 18 18.61 -0.61

9 22 18.49 3.51

s 20 19.19 0.81

11 15 19.35 -4.35

Using the Barclays share price exponential smoothing can be used to derive
forecasts.

To begin calculations, let F1 equal actual of the time series in period 1: that is F1
= Y1. Thus forecast for period 2 would be:

F2 = aY1 + (1-a)F1
= aY1 + (1- a) Y1
= Y1

So according to the exponential smoothing forecast for period two it will be


equal to the actual value of time series in period 1.

Forecast for period 2 is thus 17 however actual time series value for period 2 (Y2)
= 21. Thus, period 2 has a forecast error of 21- 17 = 4.

Exponential smoothing calculation for period 3 =

Assuming smoothing constant (a) is equal to 0.2

F3 = 02Y2 + 0.8F2 = 0.2(21) + 0.8 (17) = 17.8


Thus, the forecast for Y12 = 0.2Y11 + 0.8F11 = 0.2(15) +0.8(19.25) = 18.4

The best value for the smoothing constant is one that results in the smallest
MSE.

Graph of actual and forecast Barclays price time series with smoothing constant a = 0.2
Notice how the forecasts smooth out the irregular fluctuations of the time series, and how it lags.

Appendix J– Linear regression forecast

y = Investment (£m) X = Interest rate (%)


The table shows, how much funding in
50 10
the past Odey has received from
investors at certain rates of interest. The 100 9

graph shows a linear relationship 89 7


between the two variables. Although the
110 7.5
relationship is not perfect an equation for
this line can be derived to in order to use 130 6

it for forecasting the value of y 147 4


corresponding to each value of x, the
176 3
equation of the line is called the
estimated regression equation and can 189 2

be derived in two ways. (Anderson et al.: 200 0.5


2008)
Appendix K - Line of best fit: calculating the equation

Method 1 involves drawing the line of best fit by eye that is to pass through the
middle of all data points. Then selecting two points in order to work out the
values of the constants a and b in the equation of a straight line; Y = a + bx.
(Hennessy : 1989)

Equation of a straight line =

Y = a + bX

Whereby, a (intercept) and b (gradient) are constants and Y and X represents the
independent and dependent variable.

The constants are worked out by finding two points on the graph, for example

When x=4 , y = 153 (1)

and when x = 8, y = 94 (2)

Substituting this into the equation, Y = a +bX giving

153 = a +4b

and 94 = a + 8b

The two simultaneous equations can be solved by subtracting (2) from (1) giving
59 = 4b thus b = 14.75

Now to workout a :

If 153 = a+(4*14.75) =

 153 = a + 59
 153 – 59 = a
 a = 94

The estimated regression equation is y = 94 +14.75x

Calculating straight line equation using least squares regression line

The method used above on fitting the line of best fit by eye is prone to human
error therefore inaccurate and cannot be relied upon.

Method two known as ordinary least


squares is used position the line of best fit
so that the square of the ‘y’ distance of
each point from the line is such that the
total is minimised.

This is best done in Microsoft Excel

1) Selecte Data analysis and select regression

2) In the box besides “Input Y Range” select the independent variable


column which in this case would be investment into the fund.
3) In the “Input X range” Select the column with the Interest rate data and
select ok - The following should appear:

Interpretation

The R squared shows the accuracy of the model which is 93%. The intercept is
213.45 which tells us that a change in interest rates was forecasted to be zero,
then investment would be £213.45m. The interest rate coefficient is -14.9 tells us
that if interest rates increase by 1%, investment would decline by circa £14.8m
(Investopedia : 2011)

Appendix L – Time series case study

(Real life example) American Airlines original spare parts inventory system
used only time-series methods to forecast the demand for spare parts. This
method was slow to responds to even moderate changes in aircraft utilization let
alone major fleet expansions) American airlines then developed a PC-based
system named RAPS which uses linear regression to establish a relationship
between monthly part removals and various functions of monthly flying hours.
The computation now takes only one hour instead of the days the old system
needed. Using RAPS provided a one time savings of $7 million and a recurring
annual savings of nearly $1 million. (Anderson et.al :2008)

Reference
Anderson, D.R. , Sweeney, D.J. Williams, T.A., Martin, K (2008) An Introduction to
Management Science Quantitative Approaches to Decision Making, 12TH edition. Ohio:
Thompson South-Western.
Anbuvelan, K. (2007) Principles of Management New Delhi: Laxmi Publications Ltd

Aven,T. , Renn. O, (2010) Risk Management and Governance: Concepts, Guidelines and
Applications. London: Springer

Crundwell, F.K. (2008) Finance for engineers: Evaluation and funding of capital
projects.2nd edition. London: Springer –Verlag London Limited
Decision Craft (2010) Decision Making and Influence Diagrams, Available at:
http://www.decisioncraft.com/dmdirect/decisionmaking.htm [Date accessed: 22 Dec. 10]

Esomar, (2011) Glossary: Exponential Smoothing, Available at:


http://www.esomar.org/index.php/glossary-e.html [Date accessed: 18 Jan. 2011]
Friedland, D.K. (2000) Evidence-based medicine: a framework for clinical practice. Essex:
Mcgraw-Hill
Hennessy, L. (1989), Quantative Analysis. London: Letts
Investopedia, (2011) Regression basics for business analysts, Available at:
http://www.investopedia.com/articles/financial-theory/09/regression-analysis-basics-
business.asp [Date accessed: 20 Jan. 2011]

Investopedia (2011) Define Forecasting, Available at:


http://www.investopedia.com/terms/f/forecasting.asp [ Date accessed: 20 Jan.2011]

Investopedia, (2011) Define Time Series, Available at:


http://www.investopedia.com/terms/t/timeseries.asp [Date accessed: 20 Jan. 2011]

Investopedia (2011) Define Lagging indicator, Available at:


http://www.investopedia.com/terms/l/laggingindicator.asp [Date accessed: 21 Jan.1011]

Investopedia (2010) Define: Opportunity cost , Available at:


http://www.investopedia.com/terms/o/opportunitycost.asp [25th Dec. 10]

Lebcir, R.M. (2010) Decision Analysis, University of Hertfordshire [19th Nov. 10]

Mian, M.A (2002) Project Economics and Decision Analysis: Probabilistic models.
Oklahoma: PenWell Corporation

Myers, B. (2001), Principles of Corporate Finance, 2nd edition. London: Mgraw-Hill

Ozcan, Y.A. (2009) Quantitative Methods in Health Care Management: Techniques and
Applications. 2nd edition. San Francisco: John Wiley & Sons.

Paulton, E.C. (1994) Behavioral decision theory: a new approach. Cambridge:


Cambridge University Press

Senger, J.M. (2003) Decision Analysis in Negotiation, Department of Justice Review. Pp.1-
13.
Statistics Glossary, (2011) Define Linear Regression, Available at:
http://www.stats.gla.ac.uk/steps/glossary/time_series.html#season [Date
accessed: 21 Jan. 2011]

Statistics Glossary, (2011) Time Series data, Available at:


http://www.stats.gla.ac.uk/steps/glossary/time_series.html#season [Date
accessed: 21 Jan. 2011]
Stock Charts, (2011) Trend lines, available at:
http://stockcharts.com/school/doku.php?id=chart_school:chart_analysis:trend_lines [Date
accessed: 22 Jan. 2010]

Williard, J. (2011), Time Series Forecasting: Lecture 9. University of Hertfordshire


Myers, B. (2001), Principles of Corporate Finance, 2nd edition. London: Mgraw-Hill

Das könnte Ihnen auch gefallen