SAS Statistics Practice Exam

Question 1

What is the order of steps in both traditional statistical analyses and machine learning approaches to modeling?

Answer 1

Data exploration, variable modification, model fitting, model assessment

Question 2

Which task is best suited for deep learning as opposed to other machine learning techniques?

Answer 2

Recognizing the objects in an image or video.

Question 3

Which basic principle of statistics reflects the strength of support for results that are subject to small perturbations in input data?

Answer 3

Stability

Question 4

You are gathering data on purchase size of fountain drinks ranging from small to extra-large. What is the level of measurement for size?

Answer 4

Ordinal

Question 5

Which term has a different meaning than the other three?

Answer 5

Field

Question 6

Which terms refer to the same component of an analytical data set? (Choose 2.)

Answer 6

Object, Entity

Question 7

Which is an example of probability sampling?

Answer 7

Players are selected for teams by flipping a coin.

Question 8

Which type of sampling is used when the population is divided into groups using a categorical variable and then observations are randomly selected from these groups?

Answer 8

Stratified Sampling

Question 9

When you draw a sample of 10 balls from a bag of 100 balls of two colors using simple random sampling, what is the probability of each ball being selected?

Answer 9

1/100

Question 10

Differences in values can be meaningfully calculated for variables on which scales?

Answer 10

Ratio and Interval

Question 11

Which describes p-values in hypothesis testing?

Answer 11

A smaller p-value indicates stronger evidence against the null hypothesis.

Question 12

A lighting company claims that their LED light bulbs surpass the typical lifespan of traditional incandescent bulbs, which is 1000 hours. After examining a sample of 50 LED bulbs, the sample mean is 2200 hours, and the sample standard deviation is 200 hours. What statistical test is appropriate for determining if the new LED bulbs, on average, outlast traditional incandescent bulbs?

Answer 12

One-sample t-test

Question 13

Which measure quantifies the variability of an estimate, such as the sample mean?

Answer 13

Standard errror

Question 14

Which describes a Type II error?

Answer 14

Failing to reject the null hypothesis that is false.

Question 15

For an analysis in which the null hypothesis is known to be true, you perform a hypothesis test with α = 0.05. The analysis returns a p-value of 0.023. What does this demonstrate?

Answer 15

Type I error

Question 16

For an analysis in which the null hypothesis is known to be false, you perform a hypothesis test with α = 0.05. The analysis returns a p-value of 0.23. What does this demonstrate?

Answer 16

Type II error

Question 17

You measure patient blood pressure before and after each patient is administered a drug. How do you test the hypothesis that blood pressure remains the same before and after the drug is administered?

Answer 17

Paired t-test

Question 18

What is Pearson’s correlation coefficient (r) used for?

Answer 18

Detect collinearity

Question 19

Which of the relationships shown in the scatter plots below can be appropriately interpreted using a Pearson correlation coefficient? (Choose 2.)

Answer 19

Pearson correlation coefficients measure the strength of linear relationship, including no relationship. The Pearson correlation coefficient is not appropriate for non-linear relationships such as quadratic and cyclical.

Question 20

In a simple linear regression model where x predicts y, R2 = 0.6. What can you conclude from this information?

Answer 20

60% of the variance in y is explained by x.

Question 21

The Pearson’s correlation coefficient between x and y is 0.6. What can you conclude from this information?

Answer 21

The estimated linear association between x and y is 0.6.

Question 22

In a simple linear regression model where x predicts y, estimated β0 = 0.6. What can you conclude from this information?

Answer 22

When x=0, the predicted value of y is 0.6.

Question 23

In a simple linear regression model where x predicts y, estimated β1 = 0.6. What can you conclude from this information?

Answer 23

A one-unit change in x results in a predicted change in y of 0.6.

Question 24

Which displays a Pearson’s correlation coefficient of 0.7?

Answer 24

This graph shows a moderately strong positive linear association. /

Question 25

Which two values are divided to get the R-Square of 0.3751?

Answer 25

6593.01614, 17579

Question 26

Which describe information criteria statistics? (Choose 2.)

Answer 26

A model with a lower value shows better fit than a model with a higher value., They allow comparison between models fitted to the same data.

Question 27

Which of the following information criteria, due to its larger penalty, favors models with fewer parameters?

Answer 27

SBC

Question 28

How many parameters are estimated when fitting a linear regression model with 3 continuous predictors and one categorical predictor with 3 levels?

Answer 28

6

Question 29

Comparing two multiple linear regression models, which statistic indicates better model fit with larger values?

Answer 29

Adjusted R-square

Question 30

Performing a multiple linear regression analysis, which variable will be removed in backward elimination?

Answer 30

x5

Question 31

Which phases of predictive modeling require a known target?

Answer 31

Training a model but not scoring a new data.

Question 32

Which avoids an overly optimistic assessment of a predictive model's performance?

Answer 32

Assess the model performance on a data set not involved in fitting the model.

Question 33

In terms of accuracy and precision, which indicates overfitting in a predictive model?

Answer 33

High accuracy and low precision

Question 34

Which activities use a model to generate predictions for a data set that does not contain a target variable? (Choose 2.)

Answer 34

Model deployment, Model scoring

Question 35

Which describes explanatory modeling?

Answer 35

It is used to test an already existing set of hypotheses.

Question 36

An odds ratio of Group A to Group B is 4, with a 95% confidence interval [3,5]. What can be inferred from this?

Answer 36

Group B has 0.25 times the odds of the target event as Group A.

Question 37

To test an association between two categorical variables for statistical significance, we use which statistic?

Answer 37

Chi-square statistic

Question 38

What can a chi-square test be used for? (Choose 2.)

Answer 38

Feature selection in machine learning., Test the statistical significance of association between two categorical variables.

Question 39

In logistic regression model assessment, which indicates the best fit?

Answer 39

Maximize concordant pairs; Minimize discordant and tied pairs

Question 40

What characterizes a discordant pair? (Choose 2.)

Answer 40

Model cannot sort the event and non-event pair correctly., An observation with the event has a lower predicted probability of having the event than an observation without the event.

Question 41

Using a cutoff of 35%, which Customer IDs would be classified as non-purchasers? (Choose 2.)

Answer 41

2, 4

Question 42

A logistic regression is fit using maximum likelihood parameter estimation. How many times is the model refit when used to score new data sets?

Answer 42

Zero

Question 43

What is the default cut-off for classification allocation, in logistic regression?

Answer 43

0.5

Question 44

You are using a cutoff probability for logistic discrimination, where cases above the cutoff are allocated to class 1 (event) and cases below the cutoff are allocated to class 0 (non-event). What best describes this allocation rule?

Answer 44

The decision boundary is linear irrespective of the cutoff.

Question 45

You are using a logistic regression model to score an observation. The predicted logit for the observation is zero (0). What is the predicted probability for this observation?

Answer 45

0.5

Question 46

Which type of machine learning algorithm would you use if some of the training data has labels but most of it doesn't?

Answer 46

Semi-Supervised Learning

Question 47

A neural network with no hidden layer, one interval target, and a linear function of inputs resembles which of the following statistical models?

Answer 47

Linear regression

Question 48

What type of learning are nearest-neighbor mapping, k-means clustering, and singular value decomposition?

Answer 48

Unsupervised

Question 49

Which machine learning method is used when the cost associated with labeling is too high to allow for a full set of labeled training cases, but partial labeling information is available?

Answer 49

Semi-Supervised Learning

Question 50

A retail company wants to segment their customers into distinct groups based only on their demographic inputs. Which machine learning method is recommended?

Answer 50

Unsupervised learning

Question 51

A researcher is conducting a study on the effects of a new medication on blood pressure over a 12-week period. Due to scheduling conflicts, 25 out of 300 participants missed their week 6 check-up. The researcher believes that the likelihood of missing the check-up is related to the participants' work schedules and not their actual blood pressure levels.

Answer 51

Missing at Random (MAR)

Question 52

Which machine learning algorithm is less sensitive to unevenly scaled data, and may not necessarily require you to scale a set of features in your data?

Answer 52

Decision Tree

Question 53

Which missing data mechanism in a regression analysis would result in the slope being biased and the R-square value being reduced?

Answer 53

Missing Not at Random (MNAR)

Question 54

You are analyzing data for a predictive model, in which the target is whether a customer is likely to purchase a product. Some of the values of the variable Customer Spend are missing. You find out that this is due to random fluctuations in server traffic, occasionally observations in the database are corrupted, resulting in a small amount of missingness.

Answer 54

Missing completely at random (MCAR)

Question 55

You are analyzing data for a predictive model, in which the target is whether a customer is likely to purchase a product. Some of the values of the variable Customer Spend are missing. You find out that this is because spending data is only collected for customers who have purchased the product. What is the most likely data missingness pattern?

Answer 55

Missing not at random (MNAR)

Question 56

You are analyzing data for a predictive model, in which the target is whether a customer is likely to purchase a product. Some of the values of the variable Customer Spend are missing. You find out that this is due to a power outage that caused some of the data values to be corrupted from a specific region, resulting in a small amount of missingness in your training data.

Answer 56

Missing at random (MAR)

Question 57

Which are the most interpretable models? (Choose 2.)

Answer 57

Regression, Decision Tree

Question 58

The observed frequency for Bonus Eligible and Irregular Lot Shape is 31. What is the expected frequency for this cell?

Answer 58

13.6856 (Row Total*Column Total) / Sample Size = (93*44) / 299 = 13.6856

Question 59

In an online sale, 2500 OrderID's were generated on a given day, out of which 125 values of OrderID variable were unique. What is the cardinality of OrderID variable?

Answer 59

125

Question 60

If an observation with the target event has the same predicted probability as another observation without the target event, what type of pair is this?

Answer 60

Tie