EDA Quiz 1

You are ingesting data from multiple sources (CSV, JSON, and Parquet). Which of the following statements are correct?

• CSV files cannot handle hierarchical data structures.

• JSON files are human-readable and can handle nested objects.

• Parquet files store data in a columnar format and allow efficient compression.

• CSV files always load faster than Parquet files.

You have a 50 GB CSV file you need to ingest and analyze. Which approach(es) could be most practical?

• Use Python's built-in open() and read line by line in a loop.

• Use pandas.read_csv() without specifying chunksize.

• Use chunking in pandas (chunksize parameter) to process data in smaller batches.

• Convert the CSV into a more compressed format like Parquet and use a distributed environment (e.g., PySpark).

Which of the following summary statistics are typically useful during EDA?

• Mean, median, mode

• Standard deviation, variance

• Range, interquartile range

• Confusion matrix

When examining a dataset's distribution, which are signs that the data might be right-skewed?

• The mean is greater than the median.

• The mean is less than the median.

• A histogram shows a longer tail to the right.

• The mode is greater than the median.

You have a dataset of housing prices. You notice that some houses are extremely expensive compared to the rest. Which methods can help you identify outliers effectively?

• Box plot to detect points beyond 1.5 IQR from the quartiles.

• Z-scores to find values far from the mean.

• Dropping all data above the median price.

• Calculating the difference between max and min values.

What does a high variance in a dataset indicate?

• The data points are spread out from the mean

• The data points are closely clustered around the mean

• The dataset has a low level of variability

• The dataset has many missing values

Which transformations are commonly used to stabilize variance or reduce skew in data?

• Box-Cox transformation

• Log transformation

• Min-Max scaling

• One-hot encoding

Which of the following statements about feature engineering is correct?

• One-hot encoding is used for numerical columns to reduce skew.

• One-hot encoding transforms a categorical column into multiple binary columns.

• Binning can be used to reduce the impact of minor observation variations by grouping continuous values into bins.

• Binning should never be used because it always leads to information loss.

You have the following histogram (hypothetical description): It shows a single peak around 50. The tail extends to the right up to around 200. Most data points are between 40 and 70. Which conclusion(s) can you draw?

• The distribution is left-skewed.

• The distribution is right-skewed.

• There might be outliers on the right end.

• Data are likely symmetric around the mean.

You have a numeric feature with unknown distribution in a large dataset. Which approach(es) would you choose to quickly understand its distribution?

• Generate a histogram or density plot.

• Automatically assume it's normal without any plots.

• Create a box plot to see outliers and quartiles.

• Calculate a confusion matrix.