91Ó°ÊÓ

The article -Creep and Fatigue Characteristics of Ferrocement Slabs" (U. Ferrvcement, 1984: 399-322) reported data on \(y=\) tensile strength (MPa), \(x_{1}=\) slab thickness \((c m), x_{2}=\) load (kg), \(x_{3}=\) age at loading (days), and \(x_{4}=\) time under test (days) resulting from stress tests of \(n=9\) reinforced concrete slabs. The results of applying the BE elimination method of variable selection are summarized in the accompanying tabular format. Explain what occurred at each step of the procedure. \begin{tabular}{lccc} Step & 1 & 2 & 3 \\ Constant & \(8.496\) & \(12.670\) & \(12.989\) \\ \hline\(x_{1}\) & \(-.29\) & \(-.42\) & \(-.49\) \\ T-RATIO & \(-1.33\) & \(-2.89\) & \(-3.14\) \\ \(x_{2}\) & \(.0104\) & \(.0110\) & \(.0116\) \\ T-RATIO & \(6.30\) & \(7.40\) & \(7.33\) \\ \(x_{3}\) & \(.00159\) & & \\ T-RATIO & \(.83\) & & \\ \(x_{4}\) & \(-.023\) & \(-.023\) & \(.570\) \\ T-RATIO & \(-1.48\) & \(-1.53\) & \(92.82\) \\ 5 & 533 & \(.516\) & \\ R-5Q & \(95.81\) & \(95.10\) & \\ \hline \end{tabular}

Step by step solution

01

Initial Model with All Variables

In this step, we begin with a regression model containing all predictor variables: slab thickness \(x_1\), load \(x_2\), age at loading \(x_3\), and time under test \(x_4\). The coefficients for each variable are estimated, along with their t-ratios. The t-ratio provides the t-statistic for the test that the coefficient is different from zero. Variables with low absolute t-ratios are potential candidates for elimination as they may not significantly contribute to the model. Here, \(x_3\) has the lowest t-ratio (0.83), indicating it may not be useful.

02

Elimination of Least Significant Variable

The variable \(x_3\) is removed from the model due to its low t-ratio, suggesting it is not statistically significant in predicting tensile strength. After removing \(x_3\), the model is refitted with the remaining variables \(x_1\), \(x_2\), and \(x_4\). The t-ratios for the remaining variables \(x_1\), \(x_2\), and \(x_4\) increase, indicating potentially stronger contributions to the model. \(R^2\) has dropped slightly, from 95.81 to 95.10, less variation in \(y\) is explained after dropping \(x_3\).

03

Final Model Evaluation

In this final model, no further variables are dropped. The coefficients and t-ratios for each predictor variable \(x_1\), \(x_2\), and \(x_4\) are examined again. The variable \(x_4\) shows a massive t-ratio (92.82), indicating it is strongly related to tensile strength when considering these variables. The coefficients have settled with \(x_1\) having a substantial negative effect, while \(x_2\) and \(x_4\) have positive effects. This model is considered optimal under the backward elimination process.

Unlock Step-by-Step Solutions & Ace Your Exams!

• Full Textbook Solutions

  Get detailed explanations and key concepts

• Unlimited Al creation

  Al flashcards, explanations, exams and more...

• Ads-free access

  To over 500 millions flashcards

• Money-back guarantee

  We refund you if you fail your exam.

Over 30 million students worldwide already upgrade their learning with 91Ó°ÊÓ!

Key Concepts

These are the key concepts you need to understand to accurately answer the question.

Variable Selection

In regression analysis, selecting the right variables is crucial for building an effective predictive model. Variable selection aims to identify which factors significantly influence the output variable, in this case, tensile strength. Including too many variables can complicate the model and may lead to overfitting, where the model performs well on training data but poorly on new data.
On the other hand, omitting important variables can result in an underfitted model that doesn't capture all important influences on the outcome. The goal is to find a balance by including variables that have a meaningful impact on the response variable. In our example, the initial model considered all experimental variables like slab thickness, load, age at loading, and time under test. Variable selection helps to pinpoint which among these truly affects tensile strength by evaluating statistics such as t-ratios, which will guide whether a variable should remain in or be removed from the model.

Backward Elimination

Backward elimination is a systematic method used to simplify a regression model. Starting with all hypothesized variables included, the process evaluates the statistical significance of each. The basic approach is to remove the least significant variable, refit the model, and repeat the process until only the significant predictors remain. This results in a simpler, more interpretable model with improved generalizability.
In the exercise, backward elimination was applied to find an optimal set of predictors for tensile strength. Initially, all four variables were included. As backward elimination progressed, the age at loading variable was removed after its low t-ratio indicated that its contribution was not statistically significant. The process continued until the model could no longer dispense with any other variables without a significant drop in explanatory power, as marked by the model's R-squared values and remaining t-ratios.

Statistical Significance

Statistical significance in regression tells us whether observed relationships between predictors and the response variable are likely due to chance. It is typically assessed using t-tests on the coefficients of the predictors. If a coefficient significantly differs from zero—usually determined by a pre-set significance level like 0.05—then the variable is deemed to have a meaningful relationship with the response. This concept was key in the backward elimination method applied in the example.
During each step, t-ratios of the variables indicated how statistically significant each predictor was. Variables with low absolute t-ratios were candidates for removal since they suggested a lack of significant contribution. For instance, the age at loading was dropped due to its non-significant t-ratio in the initial model, which signposted its trivial role in predicting tensile strength.

Tensile Strength

Tensile strength is a critical property in materials science, describing how much force a material can withstand when being stretched or pulled before breaking. It provides insight into the durability and mechanical performance of materials like concrete slabs. In regression analysis, tensile strength is often modeled as a response variable influenced by several predictors.
In the context of the exercise, tensile strength was the response variable measured in megapascals (MPa). Researchers sought to understand how it is impacted by different factors, including the thickness of slabs, the load applied, and other conditions simulated during tests. Recognizing which factors significantly affect tensile strength can guide material design and assess performance in engineering fields. The backward elimination process refined these insights by identifying variables with the most influential roles in determining tensile strength.