Descriptive Statistics

Question 1

What are descriptive statistics

Answer 1

Simple descriptions of the qualities of a dataset. They can be used as a quick insight into a dataset and most of these descriptions fall into three camps: measures of the central tendency, measures of variability, and measures of frequency distributions

Question 2

What are the measures of central tendency in descriptive statistics

Answer 2

The measures of central tendency in descriptive statistics are the mean, median, and mode. They describe the central portions of the data

Question 3

What are the measures of variability in descriptive statistics

Answer 3

The measures of variability in descriptive statistics are the variance, standard deviation, range, and interquartile range. They describe the spread of the data

Question 4

What are the measures of frequency distributions in descriptive statistics

Answer 4

The measures of frequency distributions in descriptive statistics are counts and histograms. They describe the occurrences of the different observations

Question 5

What are the limitations of using descriptive statistics

Answer 5

Descriptive statistics often boil down some component of the data into a singular value, which provides a simplified insight but can be misleading and hide underlying information. It is important to understand their limitations and use them appropriately

Question 6

What is the arithmetic mean μ

Answer 6

The arithmetic mean μ is a description of the average across the population. It is calculated by taking the sum of all the samples xi and dividing it by the total number of samples n, expressed as: μ = Σ(xi) / n. Where xi is a sample from n samples in the populations , Σ is the symbol used as a “sum” operator, subscript denotes the iterator, and superscript denotes the limit

Question 7

Why is the axis along which we perform operations important

Answer 7

The axis along which we perform operations is important because taking the mean along the horizontal axis may not make sense

Question 8

What is the formula for taking the mean along the axis

Answer 8

μj = (1/n) * Σi=1 to n (xij), where:
μj is the mean value of the j-th feature/variable
n is the number of samples/observations in the dataset
xij is the value of the j-th feature/variable for the i-th sample/observation

Question 9

What are the pros of taking the arithmetic mean

Answer 9

The pros of taking the arithmetic mean are that you don’t need to sort the data, it treats all samples equally, and it is commonly used, so many are familiar with what it represents

Question 10

What are the cons of taking the arithmetic mean

Answer 10

They are sensitive to outliers, must iterate over all samples, not for categorical data

Question 11

What is the median

Answer 11

A description of the centre of the population by the middle value from the ordered list of the observed values. 50% of the observations will be above it , and 50% below

Question 12

How is the median calculated for a list x of lenght n

Answer 12

If n is odd:
Median = x(n-1)/2
if n is even:
Median = (x(n/2)-1+ x(n/2))/2
Here we index from 0, but you may see indexing from 1

Question 13

What are the pros of using the median

Answer 13

Most robust to a few outliers, identifies the middle of the dataset, when combined with the mean, we get a sense of skew in our data, and there’s no need to iterate over an entire set

Question 14

What are the cons of using the median

Answer 14

Must sort the data, which can be expensive. Different approach based on if n is odd or even. Not for categorical data

Question 15

What is the mode

Answer 15

The mode is a measure of central tendency that identifies the most frequently occurring value in the dataset

Question 16

What type of data is the mode best suited for

Answer 16

The mode is best suited for categorical data or discrete variables

Question 17

What are the pros of using the mode

Answer 17

The mode is good for categorical data and some insight into continuous data if we aggregate well, it identifies the most common observation and there is no need to sort the data

Question 18

What are the cons of using the mode

Answer 18

The con of using the mode are that it must be counted or aggregated, there may be multiple nodes, and it is not always a good reflection of the dataset as a whole

Question 19

How is the mode calculated

Answer 19

The mode is calculated by finding the value that occurs most frequently in the dataset

Question 20

What is frequency distribution

Answer 20

Frequency distribution is a way to describe the frequency of occurrence for observations within the population

Question 21

What insights can frequency distribution provide

Answer 21

Frequency distribution provides insight into questions like the distribution of animals, age group, and the concentration of values in certain areas

Question 22

What is the purpose of frequency distribution

Answer 22

To give a summary of the number of observations and the frequency of each value

Question 23

What type of data is best suited for mean

Answer 23

Continuous numerical data

Question 24

What is the purpose of frequency distributions counts

Answer 24

Frequency distribution counts provide insights into the number of times each observation occurs within a population

Question 25

What information can frequency distribution counts provide

Answer 25

Frequency distribution counts can provide information on the relative frequency of each observation, the most common observations, and the spread or dispersion of the data

Question 26

How are frequency distribution counts calculated

Answer 26

Frequency distributions counts are calculated by counting the number of times each observation appears within a population and presenting the results in a table or graph

Question 27

What is the difference between frequency distribution counts and frequency distribution percentages

Answer 27

Frequency distribution counts represents the actual number of times each observation appears within a population, while frequency distribution percentages represent the proportion of times each observation appears within a population

Question 28

Why is it important to understand frequency distribution counts

Answer 28

Understanding frequency counts can provide valuable insight into the characteristics of a population and can be used to make informed decisions based on the data

Question 29

What is a histogram in the context of frequency distribution

Answer 29

A histogram is a way of visually representing a frequency distribution for a continuous feature. It divides the observed range into fixed-sized sub-partitions called "bins" and counts the number of observations that fall within each bin

Question 30

How are values represented in a histogram

Answer 30

Values are represented on the x-axis of a histogram, while the y-axis shows the frequency (count) of observations that fall within each bin

Question 31

What is the difference between a closed and an open interval in a histogram

Answer 31

A closed interval is a histogram that includes its endpoint, while an open interval does not. In notation, a closed interval is denoted with square brackets [], while an open interval is denoted with parentheses (). E.g. [0,10],(10,20]

Question 32

What is range

Answer 32

The range is the difference between the highest and lowest values in a sample

Question 33

What does the range tell us

Answer 33

The range provides a basic description of how spread out a dataset is

Question 34

What are the pros of using the range

Answer 34

The range is simple to calculate and easy to understand

Question 35

What are the cons of using the range

Answer 35

The range is sensitive to outliers and may not provide a complete picture of the variability in the dataset

Question 36

What does population variance describe

Answer 36

Population variance describes, on average, how the samples in a population vary from the mean

Question 37

What is the formula for population variance

Answer 37

σ^2 = Σ(xi - μ)^2 / N Where xi is a sample μ is the mean of all samples N is the total number of samples in the population

Question 38

Can population variance have a negative value

Answer 38

No, it is always non-negative

Question 39

How does a larger population variance affect the variability of the samples

Answer 39

A larger population variance indicates greater variability in the samples from the mean

Question 40

What is the difference between population variance and sample variance

Answer 40

Population variance uses all individuals in the population to calculate the variance, while sample variance only uses a subset of individuals from the population

Question 41

Why is sample variance used instead of population variance in some cases

Answer 41

Sample variance is used when we don't have access to all individuals in the population and need to estimate the variance based on a sample

Question 42

What is the formula for sample variance

Answer 42

The formula for sample variance is s^2 = ∑(xi - x̄)^2 / (n-1), where xi is the value of the ith observation, x̄ is the mean of the observations, and n is the number of observations in the sample.

Question 43

Why is the denominator in the sample variance formula (n-1) instead of n

Answer 43

The denominator is (n-1) instead of n to correct for the bias that results from using a sample to estimate the population variance

Question 44

What is the unit of measurement for sample variance

Answer 44

The unit of measurement for sample variance is the square of the unit of measurement for the observation in the sample. For example, if the observations are measured in inches, the unit of the measurement for sample variance would be square inches

Question 45

What is the unit of measurement for population variance

Answer 45

The unit of measurement for population variance is square of the unit of measurement for the data. For example, if the data is measured in meters, the variance will be measured in square meters

Question 46

What is standard deviation

Answer 46

A measure of the amount of variation or dispersion of a set of values from its mean. It is calculated as the square root of the variance of the dataset, which is the average of the squared differences from the mean

Question 47

What is the difference between variance and standard deviation

Answer 47

Variance is the average of squared differences from the mean, while standard deviation is the square root of the variance

Question 48

Why is standard deviation more useful than variance

Answer 48

Standard deviation is more useful than variance because it is in the same units as the original data making it easier to interpret and compare

Question 49

What is a binary label in machine learning

Answer 49

A binary label is a label that can take on only two possible values, such as "yes" or "no", "true" or "false", or 0 or 1. Binary labels are commonly used in classification tasks where the goal is to predict a binary outcome.

Question 50

What are some examples of binary labels in machine learning?

Answer 50

Examples of binary labels include predicting whether an email is spam or not spam, whether a credit card transaction is fraudulent or not fraudulent, or whether a patient is at high risk or low risk for a particular disease.

Question 51

How do you train a model using binary labels?

Answer 51

To train a model using binary labels, you typically need a dataset that includes examples of both positive and negative outcomes. You would then use this data to train a classification algorithm, such as logistic regression or a decision tree, to predict the binary outcome based on the input features.

Question 52

What are some common evaluation metrics for models trained with binary labels?

Answer 52

Common evaluation metrics for binary classification models include accuracy, precision, recall, and F1 score. These metrics are used to measure the performance of the model in terms of its ability to correctly classify positive and negative examples.

Question 53

What is supervised learning

Answer 53

Supervised learning is a type of machine learning in which the model is trained on a labelled dataset. The labelled data consists of input features and corresponding output values, or "labels "the model is trying to predict. The goal of supervised learning is to train the model to accurately predict the output values for new, unseen input data

Question 54

What are some examples of supervised learning

Answer 54

Predicting prices of a house based on its features, classifying emails as spam or not spam, and identifying the type of flower based on its measurements

Question 55

What is unsupervised learning

Answer 55

Unsupervised learning is a type of machine learning in which the model is trained on an unlabelled dataset. The goal of unsupervised learning is to identify patterns or structure in the data without the use of predefined output values. This can be useful for tasks such as clustering, dimensionality, reduction, and anomaly detection

Question 56

What are some examples of unsupervised learning

Answer 56

Clustering similar customer profiles based on their purchasing behaviour, reducing the dimensionality of high-dimensional data for visualisation purposes, and identifying outliers in a dataset

Question 57

What is the main difference between supervised and unsupervised learning

Answer 57

The use of labelled versus unlabelled data. In supervised learning, the model is trained on labelled data to predict output values. In unsupervised learning the model is trained on unlabelled data to identify patterns or structure in data

Question 58

What is feature scaling

Answer 58

A technique used in machine learning to normalise the range of values for each feature in a dataset. This is important because individual features may have different distributions, ranges or units of measurements, which can cause issues for some machine learning algorithms

Question 59

Why is data cleaning and preparation important in machine learning

Answer 59

Data cleaning and preparation is important because machine learning algorithms require high-quality, well-structured data to produce accurate and reliable results. Dirty or unstructured data, such as missing values, error, or outliers, can significantly impact the performance of a machine learning model

Question 60

What are some common techniques used for data cleaning and preparation

Answer 60

Removing or inputting missing values, correction of errors in data entry or collection, standardising the formatting and structure of data and identifying and removing outliers

Question 61

How can you deal with multivariate data in machine learning

Answer 61

Dealing with multivariate data in machine learning requires careful consideration of the relationships between features. This may involve identifying causative variables, removing redundant or irrelevant features, and normalising the range of values for each feature using techniques such as feature scaling

Question 62

What are some best practices for data cleaning and preparation in machine learning

Answer 62

Best practices for data cleaning and preparation in machine learning include thoroughly documenting data sources and cleaning processes, validating data quality and completeness, exploring the data for patterns and outliers and using established standards and frameworks for data management and preprocessing

Question 63

What is missing data

Answer 63

Missing data refers to the absence of a value for a particular observation or variable in a dataset. This can occur for various reasons, such as data collection errors, data entry mistakes, or intentionaly omissions

Question 64

Why is missing data a problem in machine learning

Answer 64

Missing data can cause problems in machine learning because many algorithms are designed to work with complete, high-quality data. Incomplete or missing data can lead to biased inaccurate results, and can even cause the model to fail completely

Question 65

What are some techniques for inputting missing data

Answer 65

Techniques for inputting missing data include mean or median imputation regression imputation, and k-nearest neighbour imputation. Each technique has its own strengths and weaknesses and the choice of technique depends on the specific characteristics of the dataset and the research question being addressed

Question 66

What are some best practices for handling missing data in machine learning

Answer 66

Best practices for handling missing data include carefully documenting the reasons for missing data, exploring the patterns of missing data in the dataset, considering both removal and imputation techniques, validating the quality of imputed data, and performing sensitivity analyses to assess the impact of missing data on the results

Question 67

What is mean/median imputation

Answer 67

This technique involves replacing missing values with the mean or median value of the variable across all other observations. This is a simple and quick technique, but it can be biased if the missing values are not randomly distributed across the variable. Mean or median imputation is best suited for variables with relatively symmetric distribution

Question 68

What is regression imputation

Answer 68

This technique involves using a regression model to predict the missing values based on other variables in the dataset. This approach assumes that the missing values are related to other variables in the dataset, and can be useful when there is a strong correlation between variables. However, regression imputation can be computationally intensive and may require careful model selection and validation

Question 69

When is mean/median imputation best suited for

Answer 69

Mean or median imputation is best suited for variables with relatively symmetric distribution

Question 70

What is the K-nearest neighbor imputation

Answer 70

This technique involves identifying the k-neareset observations to the observation with missing values and using their values to impute the missing data. This apporach assumes that observations with similar values for other varaibles will have similar values for the variable with missing data.

Question 71

When is regression imputation best suited for

Answer 71

It is best suited when there is a strong relationship or correlation between the variable with missing values and other variables in the dataset