Analytics Exam 1

Descriptive Analysis

Investigates what is happening currently or has occurred in the past

Diagnostic

Helps understand why something happened

Predictive

Forecasts what might happen in the future

Prescriptive

Helps understand what should happen to meet goals and objectives

Example of Descriptive Analysis

What were gross sales by region for the past two years

Example of Diagnostic Analysis

Why did sales decrease in Region 1 in the prior year

Example of Predictive Analysis

What will sales be next year if we increase market share by 10%

Example of Prescriptive Analysis

What is the most cost-effective way to ship our products

Stages of the data analysis process

Plan, Analyze, Report

Stage 1 : Plan

Identifying the motivation for the analysis, determining the objective and questions to answer, and devising a strategy

Stage 2: Analyze

Data preparation, building information models, and exploring the data

Stage 3: Report

interpreting the analyses, and communicate effectively

What is motivation in data analysis

The reason the analysis is being performed. Why are we doing the analysis

What are the four risks with data analysis?

Data, Analysis, Assumptions, Biases

Data

Choosing inappropriate data, or data that are incomplete or incorrect

Analysis

Choosing an inappropriate method, or applying a data analysis method incorrectly

Assumptions

Not understanding or evaluating assumptions about the data or the results

Biases

Mental shortcuts that can affect decisions

Relational Database

A collection of logically related data that can be retrieved, manipulated, and updated to meet users’ needs.

Inner Join

Will select all of the rows from both the tables basked on the matched values

Left Join

Returns all the rows from the left table and will show any matching data from the right table

Right Join

Returns all the rows form the right table and will show any matching data from the left table

Full Join

Will return all the rows from both tables

Primary Key

A uniquely identified key

Foreign Key

If a primary key appears in another table, it is referred to as the foreign key

Attribute

The columns if the source of the data was a database

Record

Each row in a data set from a database

Fields

the individual columns in a data set

Descriptive Statistics

Help understand characteristics of data

Application of descriptive stats

Average observations in the data, the data’s shape, the distribution of the data

Correlation Analysis

Reveals relationships in data by measuring the linear relationship between two variables

How does correlation analysis work?

The measure is numerical between -1 and +1, the higher the absolute number the stronger the relationship

Skewness graph

Positive skew shows a tail off to the right, whereas the negative skew shows a tail off to the left

Positive skew

From left to right goes mode, median, mean

Negative skew

From left to right goes mean, median, mode

Coefficient of skewness

absolute value of CS > 1, high degree of skewness/ .5 < CS < 1, moderate skewness/ CS < .5, relative symmetry

Coefficient of kurtosis

If the CK > 3, data is somewhat peaked with less dispersion. If the CK < 3, data is somewhat flat with a high dispersion

What includes motivation

Opportunity, professional issues, problem solving, process and performance assessment

four types of objectives

Descriptive, diagnostic, predictive, and prescriptive

Descriptive Objectives

Designed to better understand data to answer business questions

Diagnostic Objectives

To identify a problem or issue to understand why an outcome occurred

Predictive Objectives

Focus on what may happen in the future

Prescriptive objectives

Goal is to investigate how to take advantage of future opportunities or mitigate a future risk outcome

When given the objective what questions should be asked?

Questions that address the objective, focus on a single issue, is measurable, and if the data necessary to answer it are available

Given the question, what relevant analysis should be used?

N/A

Regression analysis

Linear regression builds mathematical and statistical models to explain the relationship between a dependent variable and one or more independent variables

Linear optimization

The process of selecting values of variables that minimize or maximize some quality of interest

Data Plan Analysis

Focus on the objective, select a data strategy, select an analysis strategy, consider risks, embed controls

Measured raw data

Data created or captured by a controlled process capturing the valueu of the data. Their format can be discrete or continuous data

Nonmeasured raw data

Data often created automatically by the computer or company policy for control. These field are typically formatted as discrete data

Calculated Data

Data created when one or more fields in a particular row have any number of mathematical operators applied. These field are formatted as discrete or continuous data.

Measurement scale descriptions

Categorical, Ordinal, interval, and ratio

Categorical Data

Labeled or named data that can be sorted into groups according to specific characteristics

Ordinal Data

Ordered or ranked categorical data

Interval Data

Ordinal data that have equal and constant differences between observations and an arbitrary zero point. Examples of temperature, time, or credit score

Rational data

Interval data with a natural zero point. Economic data, such as dollars or euros

Analysis Strategies for descriptive and diagnostic analystics

Data dispersion, visualizations, correlation, calculations

Data Dispersion

Min, Max, Range, Variance, STD dev, skewness, and kurtosis

Visualizations

Bar, pie charts, histograms, and box plots

Correlation

Scatterplot

Calculations

Totals, subtotals, percent change, percent of total

Risk and embedded controls

Implement controls to reduce risks within the data and analysis strategies

Data profiling

The process of investigating data quality and structure. It has 3 components: Data quality, data structure, and deciding/informing

Table Structures

Aggregate and slice. Aggregate - calculate the total sales amount. Slice - Break the total down by region to examine the regional sales in more detail.

Quality Issues

Issues within a spreadsheet such as gender has Male and M, Date is empty, Email doesn’t have @gmail.com

Imputation

using estimated values are substituted for missing data

Data matching

A process that compares data and determines whether they describe the same entity

Star Schema

The recommended data structure for analytical databases. They consist of fact tables and dimension tables

Dimension Table

Provide context to analysis and give meaning to facts.

Fact tables

Correspond to business transactions such as orders, sales, pruchases, and payments

Cardinalities

The relationship between two tables, specifying how many rows in one table can be associated with how many rows in another table