8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 1/25
WELCOME
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 2/25
Regression Cost Model
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 3/25
Introduction Regression analysis – A statistical method by
which estimates are made of the value of avariable from a knowledge of the values of one or more other variables, and the errorsinvolved in this estimating processmeasured
Normally used in situations where therelationships between variables is notunique
Main types-
§ Simple Linear Regression Analysis
§ Multiple Linear Regression Analysis
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 4/25
Assumptions
§ The standard deviation in the error associated withthe dependent variable cost remains constantthroughout the domain
§ This error is normally distributed
§ The effect of any variable is always expressed interms of a fixed cost increase or decrease,irrespective of project size or type
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 5/25
Simple Linear RegressionAnalysis
Two-variable linear regression – describesthe relationship between two variables bycomputing a straight line through the dataobtained
Dependent variable (y) - the value to be estimated
Independent variable (x) – the factor from which theestimates are made
Constant (a)- the value of y when the independent
variable is zero
The coefficient of x (b)- The slope of the line
Expression for straight line
y = a +bx
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 6/25
§ Prediction within the range of values in the dataset is knownas interpolation
§ Prediction outside this range of the data is knownas extrapolation
a D e p e n d e n
t v a r i a b
l e
Independent variable
θ
b = tan θ
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 7/25
Steps of SLR Model
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 8/25
Specification
§ Begins with theoretical reasoning on the relationshipbetween variables
§ Form equations to represent the relationships betweenvariables
§Since the population parameters are unknown, sample isconsidered and the model is built with estimated values
Estimation
§ Lean squares estimation procedure is used most of the
time
§ Include a series of statistical tests to make sure that theestimated model is a good representation of thepostulated relationship
Steps of SLR Model
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 9/25
Validation
§ Evaluate the quality of the model
§ Evaluated on the basis of following statistics
o Coefficient of determination
o Standard error
o F ratio test- The ratio of the regression mean square to theresidual mean square
o T ratio test- The ratio of the coefficient to its standard error
Forecasting§ Forecasting should be satisfactory to the users
§ Accuracy depends on the acceptable error amount of themodel
Steps of SLR Model
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 10/25
Multiple Linear RegressionAnalysis
This aims to create a relationship with thedependant variable with several otherindependent variables.
§ Independent / Response variable - y§ Dependant/ Explanatory variables- x 1, x 2, x 3….. x n
y = a + b1 x1 + b 2 x 2 + b 3 x 3 + … bn xn + e
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 11/25
Steps of MLR Model
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 12/25
Specification
§ Begins with theoretical reasoning on the relationship
between variables§ Selecting a full set of explanatory(Independent) variables
Estimation
§ Determining the correlation coefficients between allpossible pairs.
§ Resolving multicollinearity
§ Eliminating non-significant variables one at a time until allthe remaining variables are significant.
o Use of the t-ratio- a large t-ratio is desirable.o Use of the F-ratio -Test for the significance of the overall
dependence of y on the variables (x1, x2, . . . , xn )
§
STEPS OF MLR MODEL
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 13/25
Estimation Cont’
§ Constructing a multiple linear regression modelo Making estimates of the coefficients in the regression model,
the method of least squares is used due to its simplicity.
Validation
§ Validation done before practical use in constructionindustry using another actual project.
§
Forecasting
§ If validation is a success practical use on constructionprojects
STEPS OF MLR MODEL
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 14/25
Application in ConstructionIndustry
Simple Linear Regression Analysis§ Case Study: Consider the possible sample values of
bricklayer hours and areas of brickwork from 10fictitious contracts in the following table
§
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 15/25
§Plotted scatter diagram
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 16/25
§ Scatter is caused by the factors other than area whichaffect the hours required
o Bricklayer-hours : Independent / Response variable
o Areas of brickwork : Dependent / Regression variable
§ To avoid individual judgement in constructing the line –method of least squares used
§ In fitting the regression line to a set of data, severalparameters are estimated which need to be tested forthe significance before being accepted
§ As an overall guide to the strength of association betweenthe two variables the correlation coefficient is calculated
Perfect correlation = 1
Calculated coefficient correlation = 0.998
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 17/25
§ Shows an excellent degree of correlation which cannot befound by using one variable only
§ Standard error of estimate ; anticipated differencebetween the actual values and what the regression linepredicts, should be calculated
A li i i C i
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 18/25
Multiple Linear Regression Analysis
Case Study : Obtaining a model that will estimateproductivity rates of concrete operations. For this awastewater project was observed in the North-Eastof Scotland (Project A).
The regression analysis methodology used in this study isbackward elimination, stepwise regression.
§
§ Firstly listing out the independent variables of Project Ao Type of pouro Total volume(m3)
o Number of trucks on job
o Average volume of load(m3)
o Start time
Application in ConstructionIndustry
o Average truck time(minutes)
o Number of loads
o Weather
o Concrete mix
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 19/25
§Calculating the correlation coefficients between all possible
pairs by using the in built functions of Microsoft Excel
§
Resolving multicollinearity by removing one variable (Totalvolume) out of the highly co rerated two variables ( i.e Total
volume and No. of Loads).
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 20/25
§ Estimating partial regression coefficients and thecorresponding t-statistics from the regression on actualproductivity for all eight explanatory variables.
§
§
§
§
§
§
§
§
§
§
§ Insignificant variables have small absolute values-Shouldbe eliminated
§ Carrying out two further runs, eliminating the insignificantvariables: concrete mix (t-statistic=0.97) and the starttime (t-statistic=1.72) from the regression model.
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 21/25
§ An important assumption made is the variability of thedata does not change for different levels of theresponse or explanatory variables.o This is checked by carrying out residual plots.
§
§ Constructing a multiple linear regression model for actualproductivity for a single server concrete system.
Pactual =1.31T p + 1.75V a + 0.56T n + 0.59W 0.01C t þ 0.37Ln – 6.95
Tp = Type of pour
Va = Average volume of concrete
Tn = Number of trucks on job
W = Weather
Ct = Average cycle time
Ln = Number of loads
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 22/25
§ Validation is done by using an actual concrete pours from
another wastewater project in Scotland by a different
contractor (Project B).
§
The actual productivities achieved on 32 operations observed onProject B are compared to the predicted productivities using
the derived regression model
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 23/25
Drawbacks in regression Multicollinearity§ If the explanatory variables in multiple regression are
correlated, and if the correlation coefficient (positive/negative) is high it is difficult to get their separateeffects on the dependent variable.
§ Leads to a poorly estimated partial regression coefficient.
Omitted variables§ If independent variables that have significant relationships
with the dependent variable are left out of the model,the results will not be satisfactory. E.g location, qualityetc cannot be quantified
§
Biasness of selecting independent variables§
Endogeneity§ Changes in the dependent variable cause changes in the
independent variable.
§
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 24/25
Development of RegressionModel
Development of technologies in computing,accessing, processing and storing data
§ All major statistical software packages perform leastsquares regression analysis.
§ Simple linear and multiple regression using least squares canbe done in some spreadsheet applications and on somecalculators.
§ Specialized regression software has been developed for use
in fields (survey analysis, neuro imaging).§ The Constructive Cost Model (COCOMO)- An example of an
algorithmic software cost estimation model developedusing basic regression formula.
§
8/6/2019 Regression Analysis for Cost Modelling
http://slidepdf.com/reader/full/regression-analysis-for-cost-modelling 25/25
Conclusions Regression Analysis falls under the Algorithmic Cost Model
which uses mathematical formulae linking costs/inputswith metrics to produce an estimated output.
It is used not only for estimating costs but also forforecasting productivity, time and any other parameter.
A widely used method not just in the construction Industry. When there is only one major factor affecting the response
SLR can be used When there are more than 1 major factor affecting the
respone MLR can be used There are several drawbacks and limitations in this method.
The knowledge of using Regression analysis in a specializedcost estimation software, in spread sheets and in acalculator is beneficial for the Quantity Surveyor