Quantitative Analysis: Introduction

1
Quantitative Analysis Introduction

• Geof Staniford
• Room 731
• Email g.staniford_at_livjm.ac.uk
• Telephone [phone]
• Denis Reilly
• Room 608
• Email d.reilly_at_livjm.ac.uk
• Telephone [phone]

2
Overview of the session

• Week 1 Introduction to Quantitative Analysis
• Week 2 Basic Statistics (using SPSS)
• Week 3 Statistical Testing (using SPSS)

3
Research Methods Review

• Qualitative Analysis
• Case studies
• Action research
• Thought experiments
• Non numerical
• Quantitative Analysis
• Numerical
• Experiments and surveys with numerical data
• Statistical techniques used to prove / disprove
  hypothesis

4
Quantitative Analysis and Research

• Used extensively in the natural and social
  sciences to study unpredictable complex natural
  systems
• Behaviour of people, social environment and
  nature
• Computers are predictable machines so why use
  quantitative analysis?
• Increased complexity (e.g. the Internet a vast
  collection of computers)
• The human factor
• People form an important part of the loop in the
  use of computers
• People are unpredictable, so we need to quantify
  their interaction with computers

5
Quantitative Analysis Examples

• Analysis of computer network behaviour (traffic)
• Human computer interaction
• Human perception of computers
• Use interface design and assessment
• Making computers easier to use
• Extremes of quantitative analysis
• Highly theoretical numerical study (e.g. to
  analyze computer network traffic patterns)
• Questionnaire / survey (e.g. to asses a software
  application used in an organization)
• Either way, quantitative analysis, like all
  research, calls for a plan or procedure

6
Quantitative Analysis Procedure

• The goal of quantitative analysis is to prove (or
  disprove) a theory or hypothesis using numerical
  data
• In general, this is not an easy task and calls
  for a procedure as below
• 1. State a hypothesis based on a causal
  relationship
• 2. Selection of an independent variable(s) (the
  cause) and a dependent variable(s) (the effect)
  in the relationship
• 3. Design of a controlled experiment or survey
• 4. Data collection
• 5. Data analysis using statistical methods (week
  2)
• 6. Statistical testing to provide evidence that
  proves / disproves the hypothesis (week 3)

7
Theory
Hypothesis
Selection of Variables and Measurements
Experiment / Survey Design
Survey Design Questionnaire
Experiment Manipulate Variable and Observe
Data collection
Data Analysis
Statistical Testing
8
Quantitative Analysis Procedure

• 1. The hypothesis
• Theories are very general and difficult to test
• Hypothesis considers a limited facet of a theory
• Hypothesis take the form of causal
  relationships between dependent and independent
  variables
• Goal of the experiment
• (a) Prove a causal relationship between the
  dependent and independent variables, or,
• (b) Disprove that any relationship exists (the
  so-called null hypothesis)
• Null hypothesis is usually a statement of no
  effect or no difference

9
Quantitative Analysis Procedure

• 2. Selection of dependent and independent
  variables and their scales of measurement
• Three different scales of measure
• Nominal (simply choose categories male, female)
• Ordinal (choose categories that have an ordered
  relationship small, medium, large)
• Interval (measurement scale of equal interval
  length, time, cost, age)
• Causality relationships often occur as variations
• Variation of the independent variable causes
  variation of the dependent variable
• Heavy smokers have a greater risk of poor health
  than light smokers

10
Quantitative Analysis Procedure

• 3. Experiment / survey design
• Experiments and surveys are distinguished by the
  role of the researcher
• Experiments
• The researcher can actively manipulate an aspect
  of the setting in the laboratory or out in the
  field
• In practice the independent variable or cause may
  be manipulated and the effect on the dependent
  variable then recorded
• Surveys
• The researcher does not manipulate any relevant
  aspect or variable but simply records values
• Experiments and Surveys can be combined

11
Quantitative Analysis Procedure

• 3. Experiment / survey design (contd.)
• Sampling of a subset of a population (see
  handout)
• Random or non-random sampling?
• Size of sample?
• Selection of control group as a point of
  comparison
• Mice in experimental group A are given drug X
• Mice in control group B are not
• In summary, much to do at the experiment / survey
  design stage
• The success of the analysis depends on the design
• Often several different design may be found,
  which is the best?
• Pilot studies can be used to evaluate different
  designs

12
Quantitative Analysis Procedure

• 4. Data collection
• Organization of data into a data matrix
• Rows for members of a sample
• Columns for measurements or variables for each
  member
• Use a statistical package (SPSS), spreadsheet or
  database to store data
• 5. Data analysis
• Use of basic statistical measures to make sense
  of data (week 2)
• Mean or average value, median or mid-way value
  and standard deviation
• Visualization techniques, such as frequency
  distributions, bar charts and box-plots reveal
  patterns in the data

13
Quantitative Analysis Procedure
5. Data analysis (contd.)

• Normal distributions
• Easy to deal with mean and median values are in
  the middle
• Many biological growth lifecycles are described
  by a normal distribution (plants, flowers etc.)

Frequency
Variable

• Skewed or unbalanced distributions
• Mean value is not obvious
• statistical analysis is needed to find the mean
  value

14
Quantitative Analysis Procedure

• 6. Testing the hypothesis
• Use of statistical significance tests to prove
  / disprove hypothesis (week 3)
• Tests provide court-room evidence that our
  hypothesis is true or false
• Statistics, unlike Mathematics can never give
  100
• Tests result in a probability or confidence
  factor
• Typically we may prove / disprove our hypothesis
  with a probability or confidence factor of 0.95
  (95)
• Time permitting re-running the experiment for a
  second, third, fourth etc. time with different
  samples can reinforce the results of the
  experiment

15
Relevance of Quantitative Analysis

• Quantitative analysis may be relevant to your
  research topic
• Analysis of User Interfaces and HCI both often
  use quantitative analysis techniques
• Multimedia and games often form the basis of the
  a research experiment design
• Children learning via computers is often studied
  and observed / measured using multimedia software
  or playing computer-based interactive games
• Surveys to analyze impact (usefulness) of IT in
  sectors of industry
• Computer security and network traffic
• Network traffic patterns apparent in security
  attacks (crashing web servers at 1am on New Years
  Day)