Business Analysis Statistics Test #4

Question 1

A data collection with numerous variables that lack values is used to create a prediction model. What two issues might this model have? (Select two.)

Accepted Answer

Complete case analysis means that fewer observations will be used in the model building process.

Answer

When a dataset has numerous missing values, two significant issues arise for predictive modeling. Firstly, if a complete case analysis (listwise deletion) is employed, many observations will be discarded, leading to a substantial reduction in the effective sample size available for model building. Secondly, any new cases with missing values on the input variables cannot be directly scored by the model unless specific preprocessing steps, such as imputation or deletion, are applied to handle those missing values consistently with how the model was trained.

Question 2

A logistic regression model is fitted by an analyst to determine whether a client will default on a loan or not. Agents, who each service 15-20 clients, are one of the predictors in the model. Model convergence is not achieved. The analyzed data is printed by the analyst, along with the breakdown of defaulted loans by agent. View a sample of the output below: What is the main cause of the model's inability to converge?

Accepted Answer

There is quasi-complete separation in the data.

Answer

The main cause of the logistic regression model's inability to converge, especially when examining defaulted loans by agent, is likely quasi-complete separation in the data. This occurs when one or more predictor variables (like 'Agent') perfectly or nearly perfectly predict the outcome (loan default) for a subset of observations. For instance, if all clients of a specific agent either defaulted or never defaulted, the model's algorithm cannot find a finite maximum likelihood estimate for that agent's coefficient, leading to convergence failure.

Question 3

A multiple linear regression model already exists with the addition of a non-contributing predictor variable (Pr > |t| =0.658). What will happen as a result?

Accepted Answer

An increase in R-Square

Answer

In multiple linear regression, adding any predictor variable to the model, regardless of its statistical significance, will always result in an increase or no change in the R-squared value. R-squared measures the proportion of variance in the dependent variable explained by the independent variables, and adding more variables can only account for more (or the same amount of) variance. While an insignificant predictor might not improve the model's predictive power meaningfully, it will still technically contribute to the R-squared calculation, causing it to rise.

Question 4

Choose the LOGISTIC procedure model statements that are equal. (Select two.)

Accepted Answer

Mode1 Purchase * Gender|Age|Region @1;

Answer

In SAS's LOGISTIC procedure, the `MODEL` statements `Mode1 Purchase * Gender|Age|Region @1;` and `Mode1 Purchase * Gender Age Region;` are equivalent. The `|` (bar) operator with `@1` (at-sign 1) in the first statement expands to include only the main effects of Gender, Age, and Region. The second statement explicitly lists these main effects. Both syntaxes result in a model that includes only the individual contributions of Gender, Age, and Region to predicting Purchase, without any interaction terms.

Question 5

A linear model includes the following features: a subordinate variable (y) a continuous predictor variable set of three (x1-x3) an individual category predictor variable (c1with 3 levels) Which SAS module does this model fit?

Accepted Answer

Proc glm data=sasuser.mlr;class c1;model y=c1 x1-x3 /solution;run;

Answer

The correct SAS `PROC GLM` statement for this model is `Proc glm data=sasuser.mlr;class c1;model y=c1 x1-x3 /solution;run;`. The `CLASS` statement is crucial for correctly identifying `c1` as a categorical predictor, ensuring SAS creates appropriate dummy variables for its three levels. The `MODEL` statement then specifies `y` as the dependent variable and includes the categorical `c1` along with the continuous predictors `x1` through `x3`. The `/SOLUTION` option requests the display of the parameter estimates for all model effects.

Business Analysis Practice Test

Business Analysis Practice Test

Business Analysis Statistics Test #4