Starting Your Research Journey: Research Design & Data Collection: 1
BML224: Data Analysis for Research
Learning Outcomes Aims: To discuss and contextualise the key elements of the research
design and data collec6on process
To discuss and consider the differences between different types of
data
To demonstrate how different types of data influence the type of
analysis that can be executed
Step Step 1: 1: Decide Your Decide on Research your research Topic topic
1
Cra1ing Research What? • What puzzles/intrigues me? • What do I want to know more about/understand be@er? • What are my key research quesCons?
Why? • Why will this be of enough interest to others to be published as a thesis, book, paper, guide to pracCConers or policy-‐ makers? • Can the research be jusCfied as a ‘contribuCon to knowledge?’
How – Conceptually? • What models, concepts and theories can I draw on/develop to answer my research quesCons? • How can these be brought together into a basic conceptual framework to guide my invesCgaCon?
How – Prac:cally? • What invesCgaCve styles and techniques shall I use to apply my conceptual framework (both to gather material and analyse it)? • How shall I gain access to informaCon sources/date
Step 1: Decide on Your Topic
Undertake your literature review – Read!
Step 1: Decide on Your Topic
Undertake your literature review
What is the func6on and value of your literature review?
The FuncCon of the Literature Review Demonstrates your knowledge of the research topic/problem –
originality, crea6vity, innova6on! (not merely reinven6ng the wheel)
Convinces your reader that your proposed research will make a
significant and substan6al contribu6on to the literature
Demonstrates your understanding of the theore6cal and research
issues related to your research ques6on
Shows your ability to cri6cally evaluate, integrate and synthesise
relevant literature informa6on
The FuncCon of the Literature Review Provides new theore6cal insights or develops a new model as the
conceptual framework for your research
Contextualisa6on and different perspec6ves – how has the topic
been researched before?
Helps to plan out the intended line of analysis and data
requirements
Helps you learn from other researchers’ mistakes It is expected and addresses your assessment criteria!
Assessment Criteria
Evidence of clear research aims and
objec6ves informed by background research
Managing the Literature Review Housekeeping -‐ keep a record of all your reading Note the full reference (using Harvard conven6on) You must evidence the literature review throughout (in-‐text
referencing/cita6ons)
Data and Data Sources
Principal Forms of Data: Quan%ta%ve -‐ the observa6ons or responses are expressed
numerically
Qualita%ve – use of comments, case studies or observa6ons with
responses oUen subsequently assigned to categories
Ideally aim for a balance of both quan6ta6ve and qualita6ve
methodologies/data (mixed methods)
Data and Data Sources Primary Data ata • Primary Direct Dm easurement of actual phenomena, e.g. ques6onnaires, interviews, focus groups Secondary Data
• Data collected by someone other than the user and which Primary ata generalised, filtered or modified in some way, e.g. have bDeen census data, marke6ng sta6s6cs, economy sta6s6cs, visitor numbers
Sources of Secondary Data
Sources of Secondary Data
Sources of Secondary Data
Approaches Sampling to Sampling Types of Sampling
Random Random SystemaCc StraCfied Cluster MulC-‐Stage
Non-‐Random Judgement Quota Convenience
Can you define each approach to sampling?
Step 2: Research Aims & Objectives
2
Step 2: Research Aims and ObjecCves
What is the difference between a research aim and objec6ve?
Research Aims Aims are broad statements of desired outcomes, or the general
inten6ons of the research, which 'paint a picture’/ ‘tell a story’ of your research project
Aims emphasize what is to be accomplished (not how it is to be
accomplished)
Aims address the long-‐term project outcomes, i.e. they should
reflect the aspira6ons and expecta6ons of the research topic
Research ObjecCves Research objec6ves help to explain the way in which the research
ques6on is going to be answered (they should be understandable to you and to others!)
Steps taken to answer your research ques6on/list of tasks
needed to accomplish aims
Must be sensible and precisely described and act as statements
to convey your inten6ons
Objec6ves should link to established theory and research
Research ObjecCves Research objec6ves are
usually headed by an infini6ve verb
• To iden6fy • To establish • To describe
Research objec6ves should
be met
• To determine • To es6mate • To develop • To compare • To analyse • To collect
Services
Step 3: Questionnaire Design
3
Step 3: QuesConnaire Design
What are the key elements of good ques6onnaire design?
QuesConnaire Design Layout and ordering of ques%ons Clear presenta6on and instruc6ons on about how to respond Guide the respondent clearly through the process Use pages and sec6ons in BOS Place broad and general ques6ons at the beginning of the
ques6onnaire followed by more specific ques6ons
Using funnel ques6ons
QuesConnaire Design Layout and ordering of ques%ons Numeric coding Avoid placing survey ques6ons out of order or out of context Format and layout of ques6ons – e.g. horizontal or ver6cal
presenta6on
BOS allows you to insert a variety of media into your ques6ons
(e.g. images)
QuesConnaire Design Layout and ordering of ques%ons Don’t ask two ques6ons in one – e.g. is your job interes6ng and
well paid cannot be answered with a simple yes or no
QuesConnaire Design Avoid Bias Be careful with the use of wording – the tone and wording of
ques6ons can have a significant impact on your results (e.g. might/ should/could)
Capture the Voice Even within quan6ta6ve surveys open-‐ended ques6ons provide an
opportunity to ‘capture’ the voice of the respondent
QuesConnaire Design Be Specific and Exact Unclear survey ques6ons produce answers that lack meaning and
relevance
e.g. How regularly do you watch TV – ‘regularly’ needs defining Make sure mul6ple choice ques6ons are mutually exclusive so that
a clear choice can be made
QuesConnaire Design Be Considerate Avoid intrusive ques6ons (ethics!) Consider your audience – ensure they understand your language,
terminology and the ques6on being asked
Keep ques6ons short and to the point – respondents will not
complete long and tedious surveys
Incen6ve – what do I get for comple6ng your survey?
QuesConnaire Design Choose the Right Ques%on Your choice of ques6on will influence the type of data you get Do the ques6on and answer formats provide enough robustness to
meet analysis requirements?
The type of data will influence the nature and scope of your
analysis (and ability to meet assessment criteria!)
Assessment Criteria
Clear and logical structure of the
presenta6on/poster demonstra6ng progression from basic to advanced sta6s6cal techniques referencing a specific aspect of the research process/results
Data Requirements Accurate -‐ there is no ‘systema6c bias’ in measurements Precise -‐ they are measured with sophis6cated instruments /
methodologies
Reliability -‐ they are comparable over 6me and/or spa6ally/
geographically (longitudinal analysis)
Valid -‐ they are a ‘true’ representa6on of the underlying more
complex phenomenon, e.g. ‘quality of life’ or ‘economic well-‐being’ etc.
Data Requirements Fit for purpose – provide the basis for basic descrip6ve to
advanced sta6s6cal analysis
Informed – methodologies have been drawn from a review/
cri6que of the available literature
Step 3a: Linking Data Types to Analysis
3a
Data Types
NOIR
Data Types Types of Data Nominal:
NOIR
List of categories to which objects can be a@ributed; objects
can be counted but not be measured numerically (classed as qualitaCve data)
No assumpCon on their order, only that objects in different
categories are “different”
Classed as non-‐parametric data Examples:
Which supermarket do you normally shop at?: Asda (1), Sainsburys (2), Tesco (3), Morrisons (4) Gender: Male (1), Female (2)
Data Types Types of Data Ordinal:
NOIR
List of categories but this Cme ordered or ranked; differences
are in relaCve magnitude (greater than; less than)
No assumpCon is made on the ‘distance’ between categories Classed as non-‐parame:c data
Examples:
How would you rate the quality of service provided by your mobile phone company?: 5 – Excellent to 1 – very poor
Data Types Types of Data Interval:
NOIR
ObservaCons are made on a scale comprising equal intervals
but the zero value is arbitrary
Classed as parametric data
Examples:
Fahrenheit or Celsius scale to measure temperature. Differences make sense but raCos do not (20o/10o is not twice as hot!)
Data Types Types of Data Ra:o:
NOIR
ObservaCons are made on a scale comprising equal intervals
with a true zero point
Classed as parametric data
Examples:
e.g. age, height, weight, response Cme, grade
Data Types: Further DefiniCons Further DefiniCons
QUANTITATIVE DATA
Discrete
Con:nuous
No. of people in the class
Distance traveled: 12.654 miles
INTEGER
Data Types: Further DefiniCons Types of Data -‐ Summary
NOIR NOMINAL
ORDINAL
NON-‐PARAMETRIC
INTERVAL
RATIO
PARAMETRIC
Linking Data Types to Analysis Nature of the QuesCon
NOMINAL ORDINAL Type of Data
INTERVAL RATIO
Type of Analysis DESCRIPTIVE INFERENTIAL
Linking Data Types to Analysis General Purpose
Descrip:on (only)
Specific Purpose
Summarise Data
Compare Groups
Finds Strengths of AssociaCon, Relate Variables
Type of Ques:on/ Hypothesis
Descrip:ve
Difference
Associa:onal
DescripCve StaCsCcs (e.g. mean, percentage, range)
(e.g. t-‐test, Mann Whitney)
(e.g. correla6on)
General Type of Sta:s:c
Explore Rela:onship Between Variables
[Source: Morgan, G. et al (2011), IBM SPSS for Introductory Sta6s6cs, Routledge, London, p. 6]
Linking Data Types to Analysis General Purpose
Descrip:on (only)
Specific Purpose
Summarise Data
Compare Groups
Finds Strengths of AssociaCon, Relate Variables
Type of Ques:on/ Hypothesis
Descrip:ve
Difference
Associa:onal
DescripCve StaCsCcs (e.g. mean, percentage, range)
(e.g. t-‐test, Mann Whitney)
(e.g. correla6on)
General Type of Sta:s:c
Explore Rela:onship Between Variables
[Source: Morgan, G. et al (2011), IBM SPSS for Introductory Sta6s6cs, Routledge, London, p. 6]
Linking Data Types to Analysis QuesCon: What grade do you expect to get?
Type of Data =
Type of Analysis
Linking Data Types to Analysis QuesCon: What grade do you expect to get?
Type of Data = NOMINAL
Type of Analysis
Linking Data Types to Analysis Type of Analysis
Tabular What grade do you expect to get for the module? 2010/2011
%
2011/2012
%
Grade A
6
8.5%
3
12%
Grade B
17
23.9%
16
32%
Grade C
37
52.1%
19
38%
Grade D
11
15.5%
11
22%
Grade E
0
0%
1
2%
Total
71
100%
50
100%
Linking Data Types to Analysis Type of Analysis
Graphical Expected(Grade(for(BML224(
F"(<40%):" 2%"
A"(70%+):" 6%"
D"(40,49%):" 22%"
B"(60,69%):" 32%"
C"(50,59%):" 38%"
Linking Data Types to Analysis Type of Analysis
Graphical Expected)Grade)for)BML224) 40%
35%
30%
Percentage)(%))
25%
20%
15%
10%
5%
0% A)(70%+):
B)(60169%):
C)(50159%):
Expected)Grade)
D)(40149%):
F)(<40%):
Linking Data Types to Analysis QuesCon: How confident do you feel about star%ng this module?
Type of Data =
Type of Analysis
Linking Data Types to Analysis QuesCon: How confident do you feel about star%ng this module?
Type of Data = ORDINAL
Type of Analysis
Linking Data Types to Analysis Type of Analysis
Tabular How confident are you about star:ng this module? 2010/2011
%
2011/2012
%
7 -‐ Very confident
0
0.0%
0
0.0%
6 -‐ Quite Confident
4
5.60%
1
2.0%
5 -‐ Confident
16
22.50%
10
20.0%
4 -‐ Uncertain
28
39.40%
26
52.0%
3 -‐ Anxious
14
19.70%
7
14.0%
2 -‐ Quite Anxious
4
5.60%
4
8.0%
1 -‐ Very Anxious
5
7.00%
2
4.0%
Uncertain to very anxious
51
72%
39
78%
Sample (n)
71
50
Linking Data Types to Analysis Type of Analysis
Tabular How confident are you about star:ng this module?
Average confidence level 2010/2012: 3.82 Average confidence level 2011/2012: 3.82
2010/2011
%
2011/2012
%
7 -‐ Very confident
0
0.0%
0
0.0%
6 -‐ Quite Confident
4
5.60%
1
2.0%
5 -‐ Confident
16
22.50%
10
20.0%
4 -‐ Uncertain
28
39.40%
26
52.0%
3 -‐ Anxious
14
19.70%
7
14.0%
2 -‐ Quite Anxious
4
5.60%
4
8.0%
1 -‐ Very Anxious
5
7.00%
2
4.0%
Uncertain to very anxious
51
72%
39
78%
Sample (n)
71
50
Linking Data Types to Analysis Type of Analysis
Graphical Student&Confidence&Levels&2011&
Very&Anxious:! 4%!
Quite&Anxious:! 8%!
Quite&Confident:! 2%!
Confident:! 20%! Anxious:! 14%!
Uncertain:! 52%!
Linking Data Types to Analysis QuesCon: ALtudes to Sta%s%cs
Type of Data =
Type of Analysis
Linking Data Types to Analysis QuesCon: ALtudes to Sta%s%cs
Type of Data = ORDINAL
Type of Analysis
Linking Data Types to Analysis Type of Analysis
Tabular
A`tudes Towards Sta:s:cs
This is my first ever staCsCcs class I am worried about this module If I could avoid taking this module I would I've never enjoyed maths Passing is my main goal for this module I do not see the relevance of this module
Strong Agree [5]
Agree [4]
No Opinion [3]
Disagree [2]
Strongly Disagree [1]
32%
37%
4%
14%
13%
17%
26%
18%
32%
7%
14%
32%
25%
20%
9%
10%
24%
24%
30%
13%
32%
38%
17%
10%
1%
7%
9%
24%
39%
21%
Linking Data Types to Analysis Type of Analysis
Graphical Student$Attitudes$to$Statistics$ I,do,not,see,the,relevance,of,this,module
7%$
9%$
Statement$
Passing,is,my,main,goal,for,this,module
24%$
39%$
32%$
21%$
39%$
17%$
10%$ 1%$
Strongly,Agree Agree
I've,never,enjoyed,maths
10%$
24%$
24%$
30%$
13%$
No,Opinion Disagree Strongly,Disagree
If,I,could,avoid,taking,this,module,I,would
14%$
I,am,worried,about,this,module
32%$
17%$
This,is,my,first,ever,statistics,class
25%$
25%$
18%$
32%$
0%
20%
20%$
32%$
37%$
40% Percentage$
4%$
60%
14%$
80%
9%$
7%$
13%$
100%
Linking Data Types to Analysis Type of Analysis
Graphically Attitudes$to$Statistics$
I+do+not+see+the+relevance+of+this+module
4$
Response$
Passing+is+my+main+goal+for+this+module
Ranking$Scale$ + 1+=+Strongly+Agree+ 2+=+Agree+ 3+=+No+opinon+ 4+=+Disagree+ 5+=+Strongly+Diagree+ +
2.5$
I've+never+enjoyed+maths
3.4$
If+I+could+avoid+taking+this+module+I+would
3.2$
I+am+worried+about+this+module
3$
This+is+my+first+ever+statistics+class
2.3$
0
1
2
3 Mean$Rank$
4
5
Linking Data Types to Analysis Type of Analysis
Graphically Attitudes$to$Statistics$
Plotting the mean score (rank) for each response
I+do+not+see+the+relevance+of+this+module
4$
Response$
Passing+is+my+main+goal+for+this+module
Ranking$Scale$ + 1+=+Strongly+Agree+ 2+=+Agree+ 3+=+No+opinon+ 4+=+Disagree+ 5+=+Strongly+Diagree+ +
2.5$
I've+never+enjoyed+maths
3.4$
If+I+could+avoid+taking+this+module+I+would
3.2$
I+am+worried+about+this+module
3$
This+is+my+first+ever+statistics+class
2.3$
0
1
2
3 Mean$Rank$
4
5
Linking Data Types to Analysis QuesCon: Business Turnover in 2010
Type of Data =
Type of Analysis
Linking Data Types to Analysis QuesCon: Business Turnover in 2010
Type of Data = RATIO
Type of Analysis
Linking Data Types to Analysis Type of Analysis
AnalyCcal Descrip:ve Sta:s:cs – Turnover 2010 Turnover 2010 Mean
£41,311.40
Median
£44,640.00
Mode
£44,760.00
Standard DeviaCon
£9191.0316
Linking Data Types to Analysis Type of Analysis
Graphical
Distribution of the Data
Box plot
Linking Data Types to Analysis Type of Analysis
Distribution of the Data
Graphical
Linking Data Types to Analysis Type of Analysis
Tabular Size of Business in 2010 by Category of Turnover Turnover 2010
No. of Businesses
Percentage
£0 to £9,999
0
0
£10,000 to £19,999
0
0
£20,000 to £29,999
48
16
£30,000 to £39,999
73
24
£40,000 to £49,000
118
39
£50,000 to £59,000
59
20
£60,000
2
1
300
100%
Total
Linking Data Types to Analysis Type of Analysis RECODING
RATIO
NOMINAL
Tabular Size of Business in 2010 by Category of Turnover Turnover 2010
No. of Businesses
Percentage
£0 to £9,999
0
0
£10,000 to £19,999
0
0
£20,000 to £29,999
48
16
£30,000 to £39,999
73
24
£40,000 to £49,000
118
39
£50,000 to £59,000
59
20
£60,000
2
1
300
100%
Total
Linking Data Types to Analysis Type of Analysis RECODING
NOMINAL
RATIO
Tabular Size of Business in 2010 by Category of Turnover Turnover 2010
No. of Businesses
Percentage
£0 to £9,999
0
0
£10,000 to £19,999
0
0
£20,000 to £29,999
48
16
£30,000 to £39,999
73
24
£40,000 to £49,000
118
39
£50,000 to £59,000
59
20
£60,000
2
1
300
100%
Total
Linking Data Types to Analysis Type of Analysis
Graphical Size%of%Business%by%Turnover%Category% £60,000% 1%% £50,000%to%£59,000%% 20%%
£20,000%to%£29,999%% 16%%
£30,000%to%£39,999%% 24%%
£40,000%to%£49,000%% 39%%
Linking Data Types to Analysis QuesCon: % Change in Turnover 2008-‐2010
Type of Data =
Type of Analysis
Linking Data Types to Analysis QuesCon: % Change in Turnover 2008-‐2010
Type of Data = INTERVAL
Type of Analysis Same process of analysis for RATIO data would apply
Types of Data Summary TYPE OF MEASUREMENT
QUALITATIVE NOMINAL In nominal measurement the variables consists of named categories. The categories have no mathemaCcal properCes.
QUANTITATIVE VARIABLES ORDINAL
INTERVAL
RATIO
The scores indicate only rank order in terms of size. It is not correct to calculate means on the scores.
The steps between the scores are equal in size though there is no proper zero point. The scores can be added, means calculated etc.
This is the same as ‘interval’ measurement but the scale has a proper zero point. RaCos can be calculated as a consequence
[Source: Howitt, D. and Cramer, D. (2011), Introduction to Research Methods in Psychology, Pearson, London]
Learning Outcomes At the end of this session you should be able to: Iden6fy and discuss the importance of key elements of the research
design and data collec6on process
Contextualise the research design and data collec6on process via an
ar6cula6on of the requirements of the undergraduate management project
Discuss and consider the differences between different types of data Demonstrate how different types of data influence the type of analysis
that can be executed through the design of your own survey ques6ons