Data Analysis Concepts

Question 1

What are the Data Analysis phases (in this course)?

Answer 1

AP-PASA
Ask: understand problem, goals, stakeholders - plan project
Prepare: get the data
Process: clean, organize transform
Analyze: explore, visualize, stats
Share: communicate, report, story
Act: solutions

Question 2

What is Data Analysis

Answer 2

Turning data into insights for informed action - reduce risk of wasted efforts

Question 3

What is the SMART question methodology?

Answer 3

Specific: simple, focused
Measurable: quantifiable
Action-oriented: Encourage chage
Relevant
Time-bound

Question 4

What’s the difference between a Data Analyst, Data Engineer, & Data Scientist?

Answer 4

Analyst: answers questions with existing data – SQL, spreadsheets, DB’s, BI, dashboards
Engineer: turn raw data into actionable pipelines
Scientist: creates new ways of modeling and using data

Question 5

What’s the difference between data-driven vs data-inspired decision-making?

Answer 5

Data-driven: using facts to guide strategy… requires quality & quantity… over-reliance can result in historical bias, ignoring qualitative insight
Data-inspired: adds in other sources of info - feelings/experience, difficult to measure qualities, related concepts

Question 6

Quantitative vs Qualitative data: explain differences and give examples

Answer 6

Quantitative: specific & measurable. Often gives WHAT of a problem.
* Structured interviews, surveys/polls

Qualitative: subjective or explanatory - can’t be quanitified. Often gives WHY of a problem.
* Focus groups, social media text/review analysis, in-person interviews
Powerful when combined

Question 7

Report vs Dashboard: explain differences, strengths/weaknesses

Answer 7

Report: Static, distributed periodically
+/- High level, historical
+. Quick to build, easy IF maintained
+. Static data - no cleaning
-. Continual mainteance
-. Less interactive

Dashboard: Real-time data, multiple datasets in one place
+. Dynamic, automated, interactive
+. User exploration
-. Labor-intensive design
-. Can be confusing/overwhelming (requires training)
-. More initial effort, and may need fixes
-. Potentially unclean data

Question 8

3 Types of Common Dashboard focus

Answer 8

Strategic: long term goals - highest level metrics over time frame
Operational: short-term performance and goals (most common - real-time status)
Analytical: datasets and mathematics

Question 9

Small Data vs Big Data: define and explain the differences in use

Answer 9

**Small Data: ** specific, short time-period, day-to-day decisions
- usually spreadsheets
- small/mid-size businesses
- simple to collect, store, manage, sort, visualize
- usually manageable size for analysis

**Big Data: ** larger, less-specific, longer time period, big decision
- usually database, queried
- larger businesses
- takes effort to collect, store, manager, sort, visualize
- usually needs to be broken down for analysis
- often more data than needed - challenge is to sift for gems

Question 10

What is structured thinking?

Answer 10

Process - recognize problem, organize availble info, reveal gaps/opportunities, identify options for action.

Question 11

Scope of Work & Statement of Work: Define and explain difference

Answer 11

Scope of Work: agreed upon timeline, including deliverables, milestones, and reports
Statement of Work: identifies products/services vendor or contractor will provide an organization (objectives, guidelines, deliverables, schedule, cost)

Question 12

What are the W questions to explore possible bias in data?

Answer 12

**Who: **person/organization who collected/funded
**What: **things in world the could have impacted
**Where: **origin of data
**When: **time data was created/collected
Why: motivation behind creation/collection
**How: **methods used to create/collect

Important to include context/possible bias when presenting/reporting data

Question 13

What are some tips when dealing with Executive team stakeholders?

Answer 13

• Strategic
• Headlines first
• Limited time
• Details in appendix

Question 14

1st vs 2nd vs 3rd party data sources: what’s the difference?

Answer 14

First Party: collected by individual/group themselves for own use
Second Party: collected by a group from its own audience, then sold
Third Party: collected by outside sources who didn’t collect it directly themselves - requires more checking

Question 15

Discrete vs Continuous?
Ordinal vs Nominal?
Internal vs External?

Answer 15

Discrete: whole numbers only
**Continuous: **any numeric value
Ordinal: qualitative data with set order/sequence
Nominal: qualitative data with no order/sequence
**Internal: **lives in org’s systems (more reliable, easier to collect)
**External: **lives outisde org’s systems

Question 16

Structured vs Unstructured data?

Answer 16

Structured: defined data types; usually quantitative; easy to organize/search/analyze; rows/columuns; stored in DB or warehouse
Unstructured: varied data types; usually qualitative; difficult to search; more freedom for analysis; stored in data lakes, warehouses, NoSQL DB’s; Can’t be put in rows/columns; Examples: txt msgs, soc media comments, phone call transcripts, images/audio/video

Question 17

What is Data Modeling?

Answer 17

Diagramming how data is organized/structured.
Data analysts don’t create data models, but they must be able to read/understand them.

Question 18

What are the three types of data modeling (from lowest to highest level of detail)

Answer 18

**1. Conceptual **(Business concepts): high level view of data, key entities
2. Logical (data entities): relationships, attributes, entities (not actual table/column names)
3. Physical (physical tables): specific definitions of all tables, attributes, relationships, columns, data types, ACTUAL names

Question 19

ERD vs UML

Answer 19

**ERD: **Entity Relationship Diagram - visual display of relationships between entities/database
**UML: **Unified Modeling Language - detailed diagrams that show ERD contents PLUS system behaviors and workflows (more detailed)

Question 20

Data Types:
Number
String
Boolean

Answer 20

Number: numbers only, decimal
String: text characters, punctuation
Boolean: T or F (formula uses Boolean operators AND, OR, NOT

Question 21

Wide Data vs Long Data

Answer 21

Wide Data: each subject has single row with multiple columns for attributes. Easier to compare specific attribute across different subjects. Analysts often transform Long data into Wide data for analysis/visualizations
**Long Data: **each row is one time point per subject. Each subject may have multiple rows. Versatile way to store data. More advanced, more detail on each subject.

Question 22

What is data transformation?

Answer 22

Changing data format, structure, values (adding/copying/deleting/renaming/combinging/joining datasets/reformatting).
Goal: reorganize data for easier use/analysis; improve compatibility/portability; merging datasets; ehnancing with more detail/fields, comparing data

Question 23

What are the three common types of metadata

Answer 23

Descriptive: used to identify record later on (name, ID, title, author, etc)
Structural: how organized related to other data/collections
Administrative: technical source of an asset

Question 24

What 3 elements are needed to calculate a sample size? What do they each represent?

Answer 24

Population size (total)
Confidence Level (%): how sure you are that results are representative (if you did survey again, likelihood you’d get similar results). Standard 95% or 90%
Margin of Error (+/- %): how much your results might vary from ACTUAL value. standard

Question 25

Explain differences between
- Margin of Error
- Confidence Level
- Confidence Interval

Answer 25

• Margin of Error: % variation results are from actual population result. Standard for surveys: 3%-5%
• Confidence Level: % confident that results are representative (likelihood of simialr results if repeated). Standard 90%-95%
• Confidence Interval: range of values the ACTUAL value might be at the current confidence level
  (If confidence level = 95% and margin of error is +/- 3% and value was 1,000…… “We are 95% confident that the actual value is between 9,070 and 1,030.”

Question 26

What is statistical significance?
What is statistical power?

Answer 26

statistical significance: Likelihood that outcome is due to chance (for experiment or relating two variables). Expressed as P-value between 0.0-1.0. P-value less than .05 (5%) considered statistically significant. (calculated after study).
statistical power: probability of getting meaningful (statistically significant) results from a test (determined before study). Expressed as 0.0 to 1.0. Usually power of at least 0.8 is required to consider results statistically significant.

Question 27

What is response distribution and how does it relate to margin of error?

Answer 27

Response distribution reflect the degree of variation in answers - which in turn affects margin of error. Expressed as a % where 50% is maximal variation and 0% and 100% are uniform/unanimous. (highly skewed sample reflects highly skewed population)
For this reason, the actual margin of error should be calculated for EACH QUESTION rather than for the survey as a whole.
PLAN for maximal variation (50% response distribution) but PUBLISH margin of error based on actual response distribution