Você está na página 1de 4

A Statistical Analysis on the Influence of Materials & Manufacturing Condition on the Octane Rating

Fahmid Tousif (ID-161-41630) (tous9485@vandals.uidaho.edu)

A Statistical Analysis on the Influence of Materials & Manufacturing Condition on


the Octane Rating
1) Background: Octane rating is a measure of a fuel's ability to resist knock. The octane
requirement of an engine varies with compression ratio, geometrical and mechanical
considerations and operating conditions. The higher the octane number the greater the fuels
resistance to knocking or pinging during combustion. In the US almost everybody use any sort
of motor vehicles. And octane is the most common fuel, that is used.
In this investigation of the manufacturing process in a refinery, the octane rating of a particular
petrol was measured as a function of the 3 raw materials naming material 1, material 2 and
material 3 and a variable that is characterized the manufacturing conditions. So, the purpose of
this investigation is to get the idea on how the variables are making significant differences on
the Octane Rating. (Reference: Helmut Spaeth, Mathematical Algorithms for Linear Regression,
Academic Press, 1991, ISBN 0-12-656460-4.F S Wood, The Use of Individual Effects and
Residuals in Fitting Equations to Data, Technometrics, Volume 15, 1973, pages 677-695.).
Multiple linear regression model has been used for this investigation.
2) Statistical Question: Are manufacturing condition and material 3, useful predictors for the
octane rating? And does the combination of all the variables make it significantly different?

3) Result: For 82 observations (n=82), the model consisting the variables manufacturing
condition and material 3, showed statistically significant change at an significance level of
Alpha= 0.05. The combination of all the variables, make this model a good fit, having a R
squared value of 0.90 & F-statistic value of 184.7, which is greater than the F- statistic value for
comparing two models, which has been obtained as 8.7488. Thus getting probability value of
0.0003769, implies that, the the material 3 and manufacturing condition are not significant
enough to predict the octane rating, rather a combination of all the variables are required.
Coefficients:
Estimate
Std. Error t value
Pr(>|t|)
(Intercept)
95.85315 1.224877 78.255 < 2e-16
Amount_of_material_1
-0.09282
0.005235 -17.729 < 2e-16
Amount_of_material_2
-0.1268
0.032157 -3.943
0.000176
Amount_of_material_3
-0.02538
0.013971 -1.817
0.07316
Manufacturing_condition_rating
1.967603 0.324573 6.062
4.65E-08
Table 1: Coefficient estimates for multiple regression model having all the predictors
4) Discussion: From the pair plots (Figure 1), multicolliearity has been observed to some
degree, though, a trend of stacked data points have been observed, for each of the predictors
and there are some outliers as well. From the Normal Q-Q plot (Figure2) a nice linear trend has
been observed, having very little outliers in the upper and lower tails. Thus the normality
assumptions are satisfied. Again from the residuals plot against the fittest, though many outliers
are observed in distant places, but actually, the range is really little (within +1 and -1) (Figure 3).
Again from the complete model, it has been observed that, for the 95% confidence interval, the
amount of material 3 is really not significantly associated with the octane rating (Data observed
from R output). All other three predictors are significantly associated.

A Statistical Analysis on the Influence of Materials & Manufacturing Condition on the Octane Rating
Fahmid Tousif (ID-161-41630) (tous9485@vandals.uidaho.edu)
5) Appendix:

5.1:Statistical Detailings and Analyzing Procedure:


To run the investigation, multiple regression analysis has been used. The fitted model can be
expressed as, Yi= 0+ 1X1+2X2+3X3+4X4+i
The slope estimates for all the variables (1, 2, 3, 4) and the intercept (0) have been
predicted, for the complete fitted model. The same has been done for the reduced model as
well, where the only predictors were amount of material 3 and the manufacturing condition.

The ANOVA table has been constructed using R, for both the models to investigate the Fstatistics and the P- values.
To compare this two models an F test has been used. The goal was to check, whether one of
the s is zero and if so, how it is affecting the model.
The test statistics is:

Where, the degrees of freedom can also be decomposed as follows:


n-1 = [n-(k + 1)] + (k-g) + g
If Fobs> F, k-g,n-(k+1), for =0.05, we can conclude that, the outcome variable can be significantly
explained by the complete model.
Again, For 95% confidence interval the prediction interval has also been observed on both the
complete and reduce model, to analyze the significant association of the individual predictors,
with the output model.

5.2: R code & Output (attached in separate file)

5.3: Data file (attached in separate file)

A Statistical Analysis on the Influence of Materials & Manufacturing Condition on the Octane Rating
Fahmid Tousif (ID-161-41630) (tous9485@vandals.uidaho.edu)

5.4 List of Figures:

Figure 1: Paired scatterplot of all the examined variables

A Statistical Analysis on the Influence of Materials & Manufacturing Condition on the Octane Rating
Fahmid Tousif (ID-161-41630) (tous9485@vandals.uidaho.edu)

Figure 2: Normal Q-Q Plot for the complete model

Figure 3: Residuals VS Fitted curve for the complete model

Você também pode gostar