4.0 Analysis and Findings
There are three numerical examples presented in order to evaluate the performance of the methods discussed previously. The three numerical examples are Brownlee’s Stack Loss data set, Hawkins-Bradu-Kass(1984) data set and Miller Lumber Company data set.
4.1 Brownlee’s Stack Loss Data
The data set stack loss, known as Brownlee’s Stack Loss Plant Data, contains operational data for a plant for the oxidation of ammonia to nitric acid; there are 21 observations on 4 variables. The dependent variable (Y) is Stack Loss and the independent variables are Air Flow (X1), Water Temperature (X2) and Acid Concentration (X3). Using Diagnostic Robust Generalized Potential (DRGP) method, we found that there are 4 high leverage points for these data, i.e. case no. 1, 2, 3 and 21.
Index DRGP (0.728) Index DRGP (0.728)
1 1.371 12 0.285
2 1.436 13 0.204
3 0.760 14 0.289
4 0.264 15 0.505
5 0.124 16 0.294
6 0.172 17 0.711
7 0.318 18 0.215
8 0.318 19 0.236
9 0.177 20 0.093
10 0.289 21 0.895
11 0.190
Table 4.1 : DRGP values for Stack loss data set
Using All Possible Subsets Regression Models, there are 2p -1 models, where p is the number of predictors in the full model . There are 23 -1 = 7 possible subsets for these data. Regression models with 1 variable are X1,X2 and X3 , with two variables X1X2, X1X3 and X2X3 and with three variables, X1X2X3. The results for the model selection criterion for each of the subset with classical method using OLS for both original and clean data and Robust MAD using LTS for original data are shown below:
Variable Original data with n=21 Clean data with n=17 r2 adjr2 Cp AIC r2 adjr2 Cp AIC
X1 0.846 0.838 13.336 2.912 0.743 0.726 8.038 2.23
X2 0.767 0.754 28.929 3.326 0.603 0.57...
... middle of paper ...
...or using OLS method and original data set with robust MAD using LTS method. The results showed that the classical selection criterion values are badly affected by the presence of high leverage points based on the changes of the classical selection criterion values in the clean data set compared to the original data set for all examples. The robust variable selection criterion values are not much affected by the presence of the high leverage points since the values of R2 and adjusted R2 are high indicating the good fit of the data for all the three examples.
We could conclude that the robust method is more reliable than the classical method in dealing with the variable selection criterion. Further research could be done on this study by using some other robust methods, e.g. the MM estimate, Least Median Squares (LTS) etc. which might give better and clearer results.
Accuracy: This paper demonstrates much accuracy, this is proven through the subtitles, statistics and in text citations for
Ni on the graph represents the number of failures and i is equal to 0 for the lowest dart weight failure; once 0 has been enters the value increases by 1 as weight increases. The sum on ni and i are represented in the ini column. From this graph addition data was able to be collected:
Financial market link to the real economy. Very often the markets are sensitive to many variables for example of most efficient managers with a SAT score above 1420, however according to “Chevalier and Ellison's manager characteristics model can explain only about 5% of the total variation in mutual fund returns” (James L. Davis), because the style-adjusted passive benchmark model has proved to be more efficient in work of the average mutual fund than the active one (James L. Davis).
The OLS linear regression analysis is a crucial statistics tool to estimate the relationship between variables. Usually, the estimator indicates the causality between one variable and the other (A Sykes, 1993) (e.g the product price and its demand quantity). This report will analyzes the product ‘Supa-clean’, a new cleaning agent in Cleano-max PLC, though two model: a demand function and a multivariate demand function. After analysing the estimator, the weakness and the room of improvement of this statistics tool will be discussed.
This report is composed of 2 sections. Firstly, the different proposed algorithms to estimate a
When looking at the methods in this paper, it is easy to get lost and confused amidst all
9. Tsung-Chih Wu, Chi-Hsiang Chen, Chin-Chung Li “Journal of Loss Prevention in the Process Industries,” http://www.sciencedirect.com/science/article/pii/S0950423007001362, 2008.
Table 5.6 shows the constraints of input parameters and that of responses and the goal and weights assigned to each parameter. Table 5.8 shows the values of 36 levels of combinations of process parameters that will give a high value of composite desirability (ranged from 0.79 to 0.94). Table 5.7 gives the optimal input process parametric setting for multi-response optimization.
Since model selection could not be based upon econometric tests, Frontier Economics employed combinations of the following output variables:
Compare to the “statistical generalization” in experiment and survey approach (Yin, 1994), in the case studies the method of ‘analytic generalization’ is widely used. Analytic generalization is mentioned about that a previously developed theory is used as a template with which to compare the empirical results of the case study.
When selecting vendors, the most important factors to consider are the quality, reliability, capability, financial stability, quality of materials of services provided and the risk involved working with a particular vendor.
A major application of linear programming is in portfolio selection decisions. Using linear programming it is easy to find the appropriate
The Present paper is based on secondary sources of data. The secondary data has been colledtd from various books , journals and websites.
Table 4.1 summarizes the descriptive statistics of Log of variables employed for this dissertation. This is important given that it give an idea about the dataset used
This is a peculiar manner of analysing data for predicting outcomes for the future by following a detailed protocol in order to reach maximum results. Multiple regression, is generally used to learn the relationship among variables which are independent, dependent or predictor variables. This tool is very useful in any type of business sector, which will help predict the numerous outcomes in a