Data Science with Python Certification — Questions and Answers

Question 1: What kind of data can be used with Pandas?

An ndarray
A python dict
All of these (Correct answer)
A scalar value

Correct answer: All of these

Explanation: All of these can be used as data in Pandas. Pandas provide various data structures, such as Series and DataFrame, that can hold and manipulate data. These data structures can be initialized with different types of data.

Question 2: In a box plot, what do the whiskers typically represent?

The full range (min to max) of the data
The 10th and 90th percentiles
Data within 1.5 × IQR from the quartiles (Correct answer)
The mean ± one standard deviation

Correct answer: Data within 1.5 × IQR from the quartiles

By default, box plot whiskers extend to the furthest data point still within 1.5 × IQR from Q1 or Q3; points beyond are plotted as outliers.

Question 3: What happens when you pass annot=True to sns.heatmap()?

Adds axis labels automatically
Adds a color bar annotation
Highlights the maximum value in each row
Displays the numeric value inside each cell (Correct answer)

Correct answer: Displays the numeric value inside each cell

annot=True causes sns.heatmap() to write the data value as text inside each colored cell of the heatmap.

Question 4: What is the purpose of cross-validation in supervised learning?

To get a more reliable estimate of model generalization performance (Correct answer)
To automatically tune hyperparameters to optimal values
To reduce the dimensionality of the feature space
To speed up model training on large datasets

Correct answer: To get a more reliable estimate of model generalization performance

Cross-validation repeatedly splits data into train/validation folds, providing a robust performance estimate that reduces variance compared to a single hold-out split.

Question 5: What does the `shape` attribute of a NumPy array return?

The number of axes
The total number of elements
The data type of elements
A tuple of the array's dimensions (Correct answer)

Correct answer: A tuple of the array's dimensions

The `shape` attribute returns a tuple representing the size of each dimension of the array.

Question 6: What does the `errors='coerce'` argument in `pd.to_numeric()` do?

Raises a warning but continues
Converts unparseable values to NaN instead of raising an error (Correct answer)
Rounds non-numeric strings to zero
Skips unparseable rows silently

Correct answer: Converts unparseable values to NaN instead of raising an error

errors='coerce' forces invalid parsing to produce NaN, making it easy to identify and handle dirty numeric data.

Question 7: Which of the following creates a DataFrame with a MultiIndex from a groupby result?

df.set_index(['city','dept'])
df.pivot('city','dept','value')
df.groupby(['city','dept']).mean() (Correct answer)
df.groupby(['city','dept']).mean().reset_index()

Correct answer: df.groupby(['city','dept']).mean()

Grouping by multiple columns without reset_index() yields a DataFrame with a MultiIndex on the rows.

Question 8: In agglomerative hierarchical clustering, what is the purpose of the 'linkage criterion' (e.g., 'ward', 'complete', 'average')?

To specify the distance metric used to calculate the space between individual data points.
To determine the initial number of clusters.
To set the threshold for cutting the dendrogram to form the final clusters.
To define how the distance between two clusters is measured in order to decide which clusters to merge. (Correct answer)

Correct answer: To define how the distance between two clusters is measured in order to decide which clusters to merge.

The linkage criterion in hierarchical clustering specifies how the dissimilarity between clusters is measured. [17] For example, 'complete' linkage uses the maximum distance between points in two clusters, while 'single' linkage uses the minimum distance. [6, 29] This criterion is fundamental to the algorithm's bottom-up approach of successively merging the closest pair of clusters.

Question 9: Which NumPy function returns the indices that would sort an array?

np.rank()
np.sort()
np.argsort() (Correct answer)
np.sortindex()

Correct answer: np.argsort()

np.argsort() returns the integer indices that would sort the array, not the sorted values themselves.

Question 10: What does df.astype({'age': 'int32', 'score': 'float32'}) accomplish?

Casts the specified columns to new data types to reduce memory usage (Correct answer)
Renames the columns age and score
Rounds age and score to 32 decimal places
Drops rows where age or score are non-numeric

Correct answer: Casts the specified columns to new data types to reduce memory usage

astype() with a dict changes each named column to the specified dtype, which can significantly reduce memory footprint.

Question 11: What is the bias-variance tradeoff implication of increasing model complexity on a fixed dataset?

Both bias and variance increase
Bias decreases and variance increases (Correct answer)
Bias increases and variance decreases
Both bias and variance decrease

Correct answer: Bias decreases and variance increases

More complex models fit training data more closely (lower bias) but become sensitive to noise, increasing variance and risking overfitting.

Question 12: Which method removes leading and trailing whitespace from every string in a pandas Series?

Series.str.strip() (Correct answer)
Series.str.trim()
Series.strip()
Series.str.clean()

Correct answer: Series.str.strip()

The .str accessor exposes vectorized string methods; strip() removes surrounding whitespace.

Question 13: Which of the following correctly creates a dictionary using a dictionary comprehension?

{x: x**2 for x in range(5)} (Correct answer)
[x: x**2 for x in range(5)]
{x, x**2 for x in range(5)}
(x: x**2 for x in range(5))

Correct answer: {x: x**2 for x in range(5)}

Dictionary comprehensions use curly braces with a `key: value` expression followed by a `for` clause.

Question 14: When df.groupby('region').agg({'revenue': 'sum', 'units': 'mean'}) is called, what is the index of the result?

The unique values of the 'region' column (Correct answer)
The original DataFrame's index
A RangeIndex starting at 0
A MultiIndex of (region, column)

Correct answer: The unique values of the 'region' column

After groupby().agg(), the grouping column ('region') becomes the index of the resulting DataFrame.

Question 15: Which Matplotlib method changes the range of values displayed on the x-axis?

ax.xlim()
ax.set_xlim() (Correct answer)
ax.xbound()
ax.set_xrange()

Correct answer: ax.set_xlim()

ax.set_xlim(min, max) sets the lower and upper bounds of the x-axis display range.

Question 16: What is the purpose of padding sequences when preparing text data for deep learning models?

To apply dropout regularization to embeddings
To encode words as one-hot vectors
To make all input sequences the same length for batch processing (Correct answer)
To remove rare words below a frequency threshold

Correct answer: To make all input sequences the same length for batch processing

Deep learning models require fixed-size inputs; padding (usually with zeros) extends shorter sequences to a uniform length so batches can be processed efficiently.

Question 17: Choose a non-indexed object.

SparsePanel
SparsSeries
SparseDataFrame
None of these (Correct answer)

Correct answer: None of these

In pandas, SparseDataFrame, SparseSeries, and SparsePanel are all indexed data structures, allowing label- or position-based access. Since none of the listed options is a non-indexed object, 'None of these' is the correct answer.

Question 18: You are working with a categorical feature 'product_category' which contains the values 'Electronics', 'Apparel', 'Home Goods', and 'Books'. There is no inherent order or ranking among these categories. Which encoding technique should be used to prepare this feature for a linear regression model to prevent the model from assuming a false ordinal relationship?

One-Hot Encoding, as it creates separate binary columns for each category, avoiding any implied ranking. (Correct answer)
Binary Encoding, as it is more memory-efficient than One-Hot Encoding for high-cardinality features.
Frequency Encoding, as it replaces categories with their count in the dataset.
Label Encoding, as it is computationally efficient and assigns a unique integer to each category.

Correct answer: One-Hot Encoding, as it creates separate binary columns for each category, avoiding any implied ranking.

One-Hot Encoding is the correct choice for nominal categorical data (where no order exists) when used with linear models. It creates a new binary (0 or 1) feature for each category, preventing the model from incorrectly interpreting the categories as having a quantitative relationship (e.g., that 'Books' (encoded as 3) is greater than 'Apparel' (encoded as 1)). [19, 25, 26]

Question 19: How do you select all rows where the DataFrame index is in a list [10, 20, 30]?

df.iloc[[10,20,30]]
df.loc[10:30]
df[df.index in [10,20,30]]
df.loc[[10,20,30]] (Correct answer)

Correct answer: df.loc[[10,20,30]]

df.loc[[10,20,30]] selects rows by their index labels from a list.

Question 20: Which method in pandas returns the number of missing values per column?

df.null_count()
df.missing()
df.isna().sum() (Correct answer)
df.count_nan()

Correct answer: df.isna().sum()

`df.isna()` returns a boolean DataFrame and `.sum()` aggregates True values (NaN) per column.

Question 21: Which forecasting model captures both trend and seasonality using exponential smoothing?

Holt-Winters model (Correct answer)
Simple linear regression
Random walk model
ARIMA

Correct answer: Holt-Winters model

The Holt-Winters (triple exponential smoothing) model accounts for level, trend, and seasonal components simultaneously.

Question 22: What serves as the fundamental building block for all sparse indexed data structures?

None of these
SparseArray (Correct answer)
PyArray
Sarray

Correct answer: SparseArray

Explanation: The base layer for all sparse indexed data structures is typically an array or a similar data structure that provides a contiguous block of memory. In the case of sparse indexed data structures, this base layer is often referred to as a "sparse array" or a "sparse matrix."

Question 23: What does the `thresh` parameter in DataFrame.dropna() control?

Threshold percentage of missing data
Number of rows to drop
Maximum number of NaN values allowed per column
Minimum number of non-NaN values required to keep a row (Correct answer)

Correct answer: Minimum number of non-NaN values required to keep a row

thresh=N keeps only rows/columns with at least N non-NaN values.

Question 24: When splitting data into train and test sets, why must any imputation be fit ONLY on the training set?

Because sklearn Pipelines require it
Because test data may have different dtypes
To reduce computation time during cross-validation
To prevent data leakage where test statistics influence training preprocessing (Correct answer)

Correct answer: To prevent data leakage where test statistics influence training preprocessing

Fitting on all data before splitting leaks test set statistics into training, producing overly optimistic performance estimates.

Question 25: Which of the following clustering algorithms is particularly well-suited for discovering clusters of arbitrary shapes and is robust to outliers, which it can identify as noise?

DBSCAN (Density-Based Spatial Clustering of Applications with Noise) (Correct answer)
Principal Component Analysis (PCA)
K-Means
Linear Regression

Correct answer: DBSCAN (Density-Based Spatial Clustering of Applications with Noise)

DBSCAN is a density-based clustering algorithm that excels at finding clusters of non-spherical or arbitrary shapes. [1, 10] A key advantage of DBSCAN over algorithms like K-Means is its ability to identify points that do not belong to any cluster, labeling them as noise or outliers, making it highly robust. [8]

Question 26: What value does a Pearson correlation coefficient of -1 indicate?

The variables are statistically independent
A perfect negative linear relationship (Correct answer)
A perfect positive linear relationship
No linear relationship between the two variables

Correct answer: A perfect negative linear relationship

A Pearson r of -1 means the two variables have a perfect inverse linear relationship: as one increases, the other decreases proportionally.

Question 27: What does the pandas method df.groupby('col').mean() return?

A single scalar mean value
Mean values for each unique group in 'col' (Correct answer)
The column named 'col' averaged with adjacent columns
The mean of the entire DataFrame

Correct answer: Mean values for each unique group in 'col'

groupby splits the DataFrame by unique values of 'col', then mean() computes the average of all other numeric columns within each group.

Question 28: What is the purpose of `sklearn.preprocessing.RobustScaler`?

Scales features using statistics that are robust to outliers (median and IQR) (Correct answer)
Clips values to a fixed range
Normalizes features to unit norm
Removes outlier rows before scaling

Correct answer: Scales features using statistics that are robust to outliers (median and IQR)

RobustScaler centers data on the median and scales by IQR, reducing the influence of outliers compared to StandardScaler.

Question 29: What does the 'hue' parameter do in Seaborn plots?

Maps a third variable to color encoding (Correct answer)
Adjusts the transparency of markers
Changes the plot background color
Sets the color palette globally

Correct answer: Maps a third variable to color encoding

The 'hue' parameter encodes an additional categorical or numeric variable through color differentiation.

Question 30: Which NumPy function creates an array of evenly spaced values over a specified interval?

np.zeros()
np.ones()
np.arange()
np.linspace() (Correct answer)

Correct answer: np.linspace()

np.linspace() returns evenly spaced numbers over a specified interval, including both endpoints by default.

Question 31: What does the confusion matrix entry at position [1][0] represent in binary classification?

False Negatives (Correct answer)
False Positives
True Positives
True Negatives

Correct answer: False Negatives

In a confusion matrix indexed as [actual][predicted], position [1][0] means actual positive (1) predicted as negative (0) — a false negative.

Question 32: Which of the following correctly describes 'soft clustering' in GMM versus 'hard clustering' in K-Means?

Both methods assign probabilities of membership
GMM assigns each point to one cluster; K-Means gives probabilities
GMM assigns probabilities of belonging to each cluster; K-Means assigns each point to exactly one cluster (Correct answer)
Both methods assign each point to exactly one cluster

Correct answer: GMM assigns probabilities of belonging to each cluster; K-Means assigns each point to exactly one cluster

GMM produces probabilistic (soft) assignments where each point has a probability for every cluster, unlike K-Means' hard assignments.

Question 33: Which Matplotlib function creates a horizontal bar chart?

plt.bar()
plt.hbar()
plt.barh() (Correct answer)
plt.horizontal_bar()

Correct answer: plt.barh()

plt.barh() creates a horizontal bar chart, where bars extend left or right instead of up or down.

Question 34: Which pandas method reshapes data from wide format to long format?

df.stack()
df.pivot()
df.unstack()
df.melt() (Correct answer)

Correct answer: df.melt()

df.melt() unpivots a DataFrame from wide to long format, making column headers into values.

Question 35: Which sklearn class is designed to apply different transformers to different subsets of columns in a single pipeline step?

FunctionTransformer
ColumnTransformer (Correct answer)
Pipeline
FeatureUnion

Correct answer: ColumnTransformer

ColumnTransformer allows specifying different preprocessing pipelines for different column subsets (e.g., numeric vs. categorical) within one object.

Question 36: What does `df.astype({'age': 'int32', 'salary': 'float32'})` accomplish?

Casts specified columns to smaller numeric types, reducing memory usage (Correct answer)
Creates new columns with the specified types
Validates that columns match the given types
Rounds values to fit the target type

Correct answer: Casts specified columns to smaller numeric types, reducing memory usage

Passing a dict to astype() casts each named column to its specified dtype, which can significantly reduce DataFrame memory footprint.

Question 37: Which step in the Expectation-Maximization (EM) algorithm computes the probability of each point belonging to each cluster?

Initialization step
M-step (Maximization)
E-step (Expectation) (Correct answer)
Convergence check step

Correct answer: E-step (Expectation)

The E-step computes posterior probabilities (responsibilities) of cluster membership for each data point.

Question 38: A dataset contains a 'last_login_date' column with a `datetime64[ns]` dtype. Which of the following feature engineering approaches is most effective for extracting cyclical patterns that could be useful for a predictive model?

Creating new numerical features such as 'day_of_week', 'month_of_year', and a binary 'is_weekend' flag. (Correct answer)
Converting the entire 'last_login_date' column into a single integer representing the Unix timestamp.
Applying a log transformation to the date column to normalize its distribution.
Dropping the column, as datetime objects cannot be used directly in most machine learning models.

Correct answer: Creating new numerical features such as 'day_of_week', 'month_of_year', and a binary 'is_weekend' flag.

Extracting components like the day of the week, month, or creating a flag for weekends allows a model to capture time-based patterns and seasonality (e.g., user activity might be higher on weekends or at the beginning of the month). This is a standard and highly effective technique for making datetime information useful to a model. [15, 18, 29]

Question 39: How do you parse a date column automatically when reading a CSV file with pandas?

pd.read_csv(file, datetime='date')
pd.read_csv(file, index_dates=['date'])
pd.read_csv(file, parse_dates=['date']) (Correct answer)
pd.read_csv(file, date_col='date')

Correct answer: pd.read_csv(file, parse_dates=['date'])

The `parse_dates` parameter in `pd.read_csv()` automatically converts specified columns to pandas datetime objects.

Question 40: Which technique creates new features by computing the interaction between two existing numerical features?

Polynomial feature expansion (Correct answer)
Target encoding
Feature hashing
Variance thresholding

Correct answer: Polynomial feature expansion

Polynomial feature expansion generates interaction terms and higher-degree powers of existing numerical features using sklearn's PolynomialFeatures.

Question 41: What is the key difference between NumPy's `ravel()` and `flatten()` methods when used to convert a multi-dimensional array into a 1D array?

`ravel()` returns a view of the original array whenever possible, while `flatten()` always returns a new copy. (Correct answer)
`ravel()` always returns a copy, while `flatten()` returns a view.
`flatten()` can only be used on 2D arrays, while `ravel()` works on any dimension.
There is no functional difference; they are aliases for the same operation.

Correct answer: `ravel()` returns a view of the original array whenever possible, while `flatten()` always returns a new copy.

The fundamental difference is that `flatten()` always allocates new memory and returns a copy of the data. In contrast, `ravel()` is more memory-efficient as it returns a view of the original array if possible, meaning modifications to the raveled array can affect the original array.

Question 42: What is a structured array in NumPy?

An array sorted in ascending order
A multi-dimensional array with uniform dtypes
An array with a dtype composed of named fields, like a table of records (Correct answer)
An array backed by a C struct in memory

Correct answer: An array with a dtype composed of named fields, like a table of records

Structured arrays have a compound dtype with named fields, enabling each element to hold heterogeneous data types like a database row.

Question 43: In an sklearn classification report, what does 'support' represent for each class?

The number of actual instances of that class in the test set (Correct answer)
The model's confidence score for that class
The number of support vectors for that class
The number of features used to classify that class

Correct answer: The number of actual instances of that class in the test set

Support is the count of true occurrences of each class in the test labels, reflecting the class distribution of the evaluation set.

Question 44: What is 'entity embedding' in the context of feature engineering for categorical variables?

Learning dense vector representations for categories via a neural network (Correct answer)
Hashing categories to fixed-length bit strings
Mapping categories to their mean target value
Replacing categories with their frequency of occurrence

Correct answer: Learning dense vector representations for categories via a neural network

Entity embeddings use a neural network's embedding layer to learn low-dimensional dense representations for categorical variables, capturing similarity structure.

Question 45: Which pandas function is used to reshape a DataFrame from wide format to long format?

pd.melt() (Correct answer)
pd.pivot()
pd.crosstab()
pd.stack()

Correct answer: pd.melt()

pd.melt() unpivots a DataFrame from wide to long format by converting column headers into row values.

Question 46: In a Ridge regression model, what is the effect of increasing the regularization parameter alpha?

The model becomes more complex
Coefficients grow larger to fit data more tightly
Feature selection is performed by setting some coefficients to exactly zero
Coefficients shrink toward zero, reducing overfitting (Correct answer)

Correct answer: Coefficients shrink toward zero, reducing overfitting

Higher alpha imposes stronger L2 penalty, shrinking all coefficients toward zero and reducing model variance.

Question 47: What is the effect of applying np.log1p() instead of np.log() when log-transforming a feature that may contain zeros?

It clips negative values to zero before transforming
It squares values before applying the logarithm
It computes log(1 + x), avoiding undefined results for zero values (Correct answer)
It applies base-10 logarithm instead of natural log

Correct answer: It computes log(1 + x), avoiding undefined results for zero values

np.log1p(x) computes the natural logarithm of (1 + x), which is defined at x=0 and numerically stable for small positive values.

Question 48: Which method on a scikit-learn vectorizer both learns the vocabulary and transforms the training data in one step?

transform()
fit_transform() (Correct answer)
fit()
partial_fit()

Correct answer: fit_transform()

`fit_transform()` combines `fit()` (learn vocabulary/IDF) and `transform()` (convert documents to feature matrix) in a single efficient pass over training data.

Question 49: What does the .str.contains() method return when applied to a pandas Series?

A list of matching substrings
A boolean Series indicating pattern matches (Correct answer)
A filtered DataFrame with matching rows
The count of matches per element

Correct answer: A boolean Series indicating pattern matches

.str.contains() returns a boolean Series that is True where the pattern is found.

Question 50: Which Matplotlib function creates a figure with multiple subplots and returns both the figure and axes objects?

plt.subplot()
plt.figure()
plt.subplots() (Correct answer)
plt.axes()

Correct answer: plt.subplots()

plt.subplots() creates a figure and a grid of subplots, returning (fig, axes) as a tuple.

Question 51: What does df.pivot_table(values='sales', index='region', columns='quarter', aggfunc='sum') produce?

A long-format DataFrame with one row per combination
A Series of total sales grouped by region only
A heatmap of the sales data
A cross-tabulation showing total sales per region per quarter (Correct answer)

Correct answer: A cross-tabulation showing total sales per region per quarter

pivot_table() creates a 2D summary table where rows are regions, columns are quarters, and cells contain summed sales values.

Question 52: You need to flag rows where ANY column exceeds 3 standard deviations from the mean. Which expression achieves this?

(df > 3 * df.std()).all(axis=1)
df.zscore().gt(3)
((df - df.mean()) / df.std()).abs().gt(3).any(axis=1) (Correct answer)
df.gt(df.mean() + 3)

Correct answer: ((df - df.mean()) / df.std()).abs().gt(3).any(axis=1)

Computing Z-scores, taking absolute values, checking gt(3), and using any(axis=1) identifies rows with at least one outlier column.

Question 53: Which pandas method fills forward-propagates missing values along a column?

interpolate(method='linear')
fillna(method='bfill')
fillna(method='ffill') (Correct answer)
dropna(how='all')

Correct answer: fillna(method='ffill')

ffill (forward fill) propagates the last valid observation forward to replace NaN values.

Question 54: What does the following Python code produce? a, b = 0, 1 while b < 10: print(b, end=' '); a, b = b, a+b

1 1 2 3 5
1 2 3 5 8
1 1 2 3 5 8
1 1 2 3 5 8 1 3 (Correct answer)

Correct answer: 1 1 2 3 5 8 1 3

This prints the Fibonacci sequence while b < 10. Tracing b: 1, 1, 2, 3, 5, 8, then b becomes 13 which fails b < 10 and stops — so the output is '1 1 2 3 5 8'. ⚠ The stored key ('1 1 2 3 5 8 1 3') is wrong; the correct output is '1 1 2 3 5 8'.

Question 55: In the context of NLP, what is a word embedding?

A binary feature indicating presence or absence of a word
A dense continuous vector representation that captures semantic relationships between words (Correct answer)
A statistical count of how often a word appears in a document
A one-hot encoded vector where each dimension represents a vocabulary word

Correct answer: A dense continuous vector representation that captures semantic relationships between words

Word embeddings (e.g., Word2Vec, GloVe) map words to dense low-dimensional vectors where semantically similar words are geometrically close, unlike sparse one-hot encodings.

Question 56: Which NumPy operation performs element-wise multiplication of two arrays (NOT matrix multiplication)?

A @ B
np.dot(A, B)
A * B (Correct answer)
np.cross(A, B)

Correct answer: A * B

The * operator performs element-wise (Hadamard) multiplication; np.dot and @ perform matrix multiplication.

Question 57: What will `'data'[::-1]` return in Python?

'data'
Raises IndexError
'atad' (Correct answer)
'dta'

Correct answer: 'atad'

The slice `[::-1]` reverses a string by stepping backwards through all characters.

Question 58: Which evaluation metric is most appropriate when false negatives are more costly than false positives, such as in cancer detection?

Specificity
F1 Score
Precision
Recall (Correct answer)

Correct answer: Recall

Recall (sensitivity) measures the proportion of actual positives correctly identified, minimizing missed cases (false negatives).

Question 59: What does `np.pad(a, pad_width=1, mode='constant', constant_values=0)` do to a 2-D array?

Normalizes each row to sum to 1
Clips values outside the range [0, 1]
Adds a border of zeros one element wide around the array (Correct answer)
Repeats edge values once on each side

Correct answer: Adds a border of zeros one element wide around the array

`np.pad` with `mode='constant'` and `constant_values=0` adds one row/column of zeros to every edge of the 2-D array.

Question 60: What does the Area Under the ROC Curve (AUC) represent for a classification model?

The model's overall accuracy at the default classification threshold of 0.5.
The total number of correct predictions made by the model.
The probability that the model will rank a randomly chosen positive instance higher than a randomly chosen negative instance. (Correct answer)
The trade-off point where precision and recall are perfectly balanced.

Correct answer: The probability that the model will rank a randomly chosen positive instance higher than a randomly chosen negative instance.

The AUC score provides a single number summarizing the performance of a classifier across all possible classification thresholds. Its probabilistic interpretation is that it measures the likelihood that the model will assign a higher score (probability) to a randomly selected positive example than to a randomly selected negative example. An AUC of 1.0 represents a perfect model, while 0.5 represents a model with no discriminative ability.

Data Science with Python Certification

A comprehensive certification that validates proficiency in Python-based data science skills including data manipulation with Pandas/NumPy, statistical analysis, machine learning, natural language processing, and time series forecasting.

Questions

90 min

Time Limit

70%

To Pass

Checking for saved progress...

Exam Rules

You can skip questions and return to them later
Flag questions for review before submitting
No feedback shown until you submit the entire exam
Unanswered questions count as wrong — answer everything
10 pretest questions are mixed in and don't affect your score
Timer auto-submits when time runs out
Your progress is auto-saved every 30 seconds