## Analysis of results: quantitative and qualitative

### Quantitatative data analysis

One of the biggest mistakes that people make about quantitative data analysis:

• I don't have to concern myself with how I'm going to analyse my survey data until after I've collected my data. I'll leave thinking about it until then, because it doesn't impinge on how I collect my data

You should be fully aware of what techniques you will apply at a fairly early stage:

• you cannot apply just any technique to any variable as techniques must match the types of variables you have created and you must be fully conversant with the ways in which different types of variable are classified
• the size and nature of your sample are likely to impose limitations on the kinds of techniques you can use (see Chapter 8, 'Kind of analysis')

#### Types of Variable (page 317; Table 15.1)

• Interval/ratio variables - where the distances between the categories are identical across the range
• Ordinal variables - whose categories can be rank ordered but the distances between the categories are not equal across the range
• Nominal variables - whose categories cannot be rank ordered; also known as categorical
• Dichotomous variables - containing data that have only two categories

#### Deciding How to categorize a variable (Fig.15.1)

Descriptive and inferential statistics

Descriptive statistics:

• methods used to describe general trends and patterns in a data set
-uses measures of central tendency and dispersion
• provide a summary of the data collected
• associated with inductive approaches and Tukey's (1977) Exploratory Data Analysis

Inferential (confirmatory) statistics:

• methods used to test hypotheses about relationships or differences (estimates, generalizations and predictions) in populations based on measurements made on samples
- uses measures such as Pearson's r and tests such as Chi-square and analysis of variance
• provide inferences that extend beyond the data collected
• associated with deductive approaches and correlation, bi- and multivariate analysis

#### Using SPSS

You will need to discover whether your institution supports SPSS, many actually sell registered versions of SPSS at a discount price for student use.

www.policymagic.org/spss_tutor.htm - links to SPSS tutorials and resources hosted by John G. McNutt

#### Checklist on doing and writing up quantitative data analysis

• Have you made sure that you have presented only analyses that are relevant to your research questions?
• Have you made sure that you have taken into account the nature of the variable(s) being analysed when using a particular technique (i.e. whether nominal, ordinal, interval/ratio, or dichotomous)?
• Have you used the most appropriate and powerful techniques for answering your research questions?
• If your sample has not been randomly selected, have you made sure that you have not made inferences about a population (or at least, if you have done so, have you outlined the limitations of making such an inference?)?
• If your data are based on a cross-sectional design, have you resisted making unsustainable inferences about causality?
• Have you remembered to code any missing data?
• Have you commented on all the analyses you present?
• Have you gone beyond univariate analysis and conducted at least some bivariate analyses?
• If you have used a Likert scale with reversed items, have you remembered to reverse the coding of them?

http://onlinestatbook.com/rvls.html - Rice Virtual Lab in Statistics, excellent site from Rice University Texas includes:

• An online statistics book with links to other statistics resources on the web
• Examples of real data with analyses and interpretation
• Some basic statistical analysis tools.

www.42explore.com/statistics.htm - links and activities on probability and introductory statistics from 42Explore

### Qualitative data analysis (chapter 24)

#### General strategies of qualitative data analysis:

• analytic induction*
• grounded theory*

(*Note - their iterative nature means they can also be viewed as strategies of data collection, see Key concept 24.3)

#### Strauss and Corbin's Classification of coding in grounded theory (Key concept 24.2)

Open coding

• 'breaking down, examining, comparing, conceptualizing, and categorizing data' (Strauss and Corbin, 1990: 61)
• this process yields concepts, which are later grouped and turned into categories

Axial coding

• 'data are put back together in new ways after open coding, by making connections between categories' (1990: 96)
• this is done by linking codes to contexts, consequences, patterns of interaction and to causes

Selective coding

• 'selecting the core category, systematically relating it to other categories, validating those relationships, and filling in categories that need further refinement and development' (1990: 116)
• a core category is the central issue or focus around which all other categories are integrated

#### Steps and considerations in coding (page 531)

Lofland and Lofland (1995) give the following considerations:

• Of what general category is this item of data an instance?
• What does this item of data represent?
• What is this item of data about?
• Of what topic is this item of data an instance?
• What question about a topic does this item of data suggest?
• What sort of answer to a question about a topic does this item of data imply?
• What is happening here?
• What are people doing?
• What do people say they are doing?
• What kind of event is going on?

Steps in coding:

• Code as soon as possible
• Read through your initial set of transcripts, field notes, documents, etc.
• Do it again
• Consider more general theoretical ideas in relation to codes and data
• Any one item or slice of data can and often should be coded in more than one way
• Do not worry about generating what seem to be too many codes
• Keep coding in perspective

#### Other methods for qualitative data analysis:

Narrative analysis:

• Narratives should be viewed in terms of the functions that the narrative serves for the teller
• The aim of narrative interviews is to elicit interviewees' reconstructed accounts of connections between events and between events and contexts
• For the management researcher, narrative analysis can prove extremely helpful in:
• providing a springboard for understanding what Weick (1995) has termed 'organizational sensemaking'
• understanding the internal politics of organizations (see Research in focus 5.4)

#### Using NVivo (Ch.25)

• Some readers will decide it is not for them and that the tried-and-tested scissors and paste will do the trick
• On the other hand, the software warrants serious consideration because of its power and flexibility
• See pages 539-42 for arguments for and against the use of CAQDAS

Free copy of the booklet, Getting Started in NVivo from QSR International at: http://www.qsrinternational.com/nvivo/free-nvivo-resources/getting-started.

Checklist on content analysis of documents

Can you answer the following questions?

• Who produced the document?
• Why was the document produced?
• Was the person or group that produced the document in a position to write authoritatively about the subject or issue?
• Is the material genuine?
• Did the person or group have an axe to grind and if so can you identify a particular slant?
• Is the document typical of its kind and if not is it possible to establish how untypical it is and it what ways?
• Is the meaning of the document clear?
• Can you corroborate the events or accounts presented in the document?
• Are there different interpretations of the document from the one you offer and if so what are they and why have you discounted them?

### Exercise: content analysis

Go to the websites for three large companies in the same or different industrial/commercial sectors. Examples might be:

www.sony.net

www.bp.com

www.co-op.co.uk

1

##### Question

Examine the code of conduct/ethics/business of your chosen three companies and perform a content analysis on them in a way you see as most appropriate.

There are a number of ways of approaching this exercise. The company codes could be analysed by words, themes or subjects. Whichever is chosen, coding will play an important part both in the design of the schedule and the manual.

In the example above, content analysis will involve the design of a coding schedule and a coding manual. The coding schedule will be used to analyse the data into meaningful phrases/words for analysis and the coding manual will be a guide to the researcher for coding the phrases and words.

1. The first task is to break down the company codes into words or phrases which can be used to describe the information across all three cases (Information).
1. The second task is to number the words or phrases chronologically (Number).
1. The third task is to allocate a code (Code) which describes the data.
1. From this, SPSS, or another similar data analysis package, could be used to analyse the data. Students might wish to perform the SPSS analysis themselves to try and content analyse the company codes and look for similarities and differences. In this case, all three companies will be analysed using the same schedule and manual and, although the Codes might differ from company to company, the Number and Information should not. The company codes thus disaggregated, could then be input into SPSS. In this case, there would be three rows (Companies) of data in the SPSS input sheet and twelve columns representing the coded information. The Code numbers would be input into the columns, which would represent the Information. The columns would be headed with abbreviations representing the Information description.

The four stages are exemplified as follows:

The Coding Schedule. In this case the coding schedule is based on words and themes. An example is as follows:

Number Information Code
1 Company name 1
2 Location 3
3 Type of company 22
4 Size by turnover 3
5 Comprehensiveness/coverage of code:
5a Environment 2
5b Suppliers code 4
5c Customer promises 5
5d Ethical behaviour 3
5e Employee relations 2
6 Who is committed to code 1
7 Language of code:
7a Tense of verbs 1
7b Will do rather than 'believe' 1
7c etc

The Coding Manual. In this case the coding manual used could be as follows:

Number Information Code
1 Company name 1 = Company 1, 2 = Company 2, 3 = Company 3
2 Location 1 = Europe, 2 = Americas, 3 = Asia, 4 = Australasia
3 Type of company Use the SIC (Standard Industrial Classification) number
4 Size by turnover 1 =, 0 - 2\$billion, 2 = \$2-5\$billion, etc
5 Comprehensiveness/coverage of code:
5a Environment 1 = one element of the environment, 2 = two elements of the environment, etc.
5b Suppliers code 1 = one mention of suppliers, 2 = two mentions of suppliers, etc.
5c Customer promises 1 = one promise, 2 = two promises, 3 = three promises, etc.
5d Ethical behaviour 1 = one ethical action, 2 = two ethical actions, etc.
5e Employee relations 1 = one mention of employee relations, 2 = two mentions of employee relations, etc.
6 Who is committed to code 1 = Chairman and CEO, 2 = Board, 3 = stakeholders, etc.
7 Language of code:
7a Tense of verbs 1 = present, 2 = past, 3 = future
7b Will do rather than 'believe' 1 = current action, 2 = future action
7c etc.

Input into SPSS

The first line of the data sheet in SPSS would be as follows (assuming the above data is for company one only):

Comp Loc'n Type Turn'r Envir't Supp'r Custo'r Eth bev Empl Who Tense Will
1 3 22 3 2 4 5 3 2 1 1 1

### Exercise: quantitative data analysis

2

##### Question

a) Given the following data calculate the arithmetic mean.

Number of bottles of wine consumed per month

Frequency (f)

1-5

7

6-10

11

11-15

17

16-20

13

21-25

8

26-35

10

36 and over

5

b) What are the features of the mean?

a) First explain what an arithmetic mean is as opposed to a geometric mean i.e. the sum of all values in a series and dividing by the total number of items in the series.

This exercise is grouped data rather than ungrouped data, so the exact values of each item are not known. This makes it necessary to approximate somewhat by taking the midpoint of each class interval to represent the whole class. If an assumed mean is chosen it should be the midpoint of the class and any deviations (d) from the assumed mean (A) are measured from the midpoint of the class in question.

In order to answer the question the mid point of the class has to be calculated. This value is then multiplied by the frequency of occurrence to get the correct number of mean values. The mean can then be calculated by summing all these values and dividing by the sum of the frequencies.

No of bottles of wine consumed per month

Frequency

Cumulative Frequency (f)

0.5 - 5.5

7

7

5.5 - 10.5

11

18

10.5-15.5

17

35

15.5-20.5

13

48

20.5-25.5

8

56

25.5-35.5

10

66

35.5 - 45.5

5

71

Sum = 71

b) Features of the mean:

easy to understand and calculate and most common average

uses every value in the distribution, therefore, mathematically exact can be used for further data processing

it can be determined if only the total value of the items and the number of items are known, without knowing individual values

it can be distorted by extreme values in the distribution

for a discrete distribution the mean may be an "impossible" figure i.e. 17.68 when the value in the distribution are whole numbers

3

##### Question

a) Given the following data calculate the median and mode:

Number of bottles of wine consumed per month

Frequency (f)

0.5-5.5

7

5.5-10.5

11

10.5-15.5

17

15.5-20.5

13

20.5-25.5

8

25.5-35.5

10

35.5-45.5

5

b) What are the features of the median and mode?

a) In order to locate the median it is entered to calculate the cumulative frequency.

No of bottles of wine consumed per month

Frequency

Cumulative Frequency (f)

0.5 - 5.5

7

7

5.5 - 10.5

11

18

10.5-15.5

17

35

15.5-20.5

13

48

20.5-25.5

8

56

25.5-35.5

10

66

35.5 - 45.5

5

71

Sum = 71

The distribution has an odd number of values. The median is therefore the value of:

By looking at the data of cumulative frequencies the 36th item is between 15.5 and 20.5 values:

The median is 20.5 + 2/8 x 5 = 15.88

The mode for grouped data is the class with the highest frequency provided that all class intervals are equal. Otherwise frequency density must be considered.

The modal class is therefore 10.5 - 15.5.

The modal value is:

L = Lower limit of modal class

Cl = Class Interval

f l, fo. f2 = frequencies of modal class, next lower class and next higher class respectively.

b) Features of the median and mode

Median

• it is a measure of rank or position. Half the items in the series will have a value greater than or equal to the median and half less than or equal to the median
• it is easy to understand
• it is unaffected by the presence of extreme items in the distribution
• if found directly from (ungrouped data) it will be the same as an actual item in the distribution
• it may be found when the values of all the items are known, provided that values of middle items and the total number of items are known
• ranking the items can be tedious
• the median cannot be used for further mathematical processing
• it may not be representative if there are few items

Mode

• for discrete data it is an actual single value
• for continuous data it is the point of highest frequency density
• it is easy to understand
• extreme items do not affect its value
• it can be estimated from incomplete data
• it may not be unique or clearly defined
• it cannot be used for further mathematical processing

4

##### Question

What are the criteria for the use of the most appropriate average (mean, median and mode in any particular case)?

There is no ideal situation but these are some criteria:

To determine what would result from an equal situation use the mean e g. to determine per capita consumption of beer

If position or ranking is involved, use the median, which gives the halfway value e.g. a salesman interested if his monthly sales places him in the upper or lower half of the sales team will need to compare his sales figure with the median sales figure

Where the most typical value is required use the mode e.g. a shoe shop manager might want to know the average size of men' or women's shoes. For sales planning it will be the mode that he requires as it will tell him the most common size of shoe.

### Exercise: using SPSS for Windows

A small, explorative survey (20 responses) was done in a supermarket car park to determine what factors were important to buyers when buying a car. The 4 most important factors identified by the buyers were price, service interval, an assessment of quality and value for money.

Buyers were given a questionnaire which had a 5-point interval scale on which they could rate their preference. 1 on the scale meant very important and 5 very unimportant. In between points on the scale were intended to allow for degrees of preference between the polar extremes. The results were as follows:

Respondent

Price

Service interval

Quality

Value for money

1

1

2

2

2

2

2

2

3

1

3

1

3

3

2

4

2

1

4

2

5

1

2

3

2

6

1

3

3

1

7

2

3

2

3

8

1

3

2

1

9

1

1

2

1

10

2

1

2

2

11

1

2

3

3

12

2

3

3

2

13

1

3

1

2

14

1

3

2

2

15

1

3

2

4

16

1

3

2

2

17

2

2

2

1

18

1

3

2

2

19

1

2

3

2

20

1

3

2

1

5

##### Question 1

Given the data, use the SPSS package to:

a) Describe the data

b) Draw inferences from the data

Input data into SPSS as below:

1. Descriptive statistics.

Assume the data is interval, normally distributed, homogenous and parametric (NB. with this small sample the data, it probably would not be normally distributed, but assume so for the purposes of the exercise.)

The permissible descriptive statistics for interval data are:

Range, Mean, Standard Deviation and Product Moment Correlation

For the Product Moment Correlation, the data is from one sample so use a Pearson's Correlation

NB. to analyse the data in SPSS use the AnalySe button on the toolbar

• AnalySe
• Descriptive Statistics
• Descriptives/Descriptives Options

Use the same AnalySe box for the other computations but use the test menus

1. Inferential statistics

Use a t test one sample

To test if there is any significant difference between the means of the 4 variables:

• Set up a number of null hypotheses for the different variables
• Use a Paired t test. Choose a significance level.

Interpretation of the paired T test

1. Compare parameters i.e. mean, standard deviation or standard errors to see if there are obvious differences between the groups
2. Observe if the t value is significant or not. NB it should not be if the 3 measures in 1 are close

If the T value is 0.50 or above it could be significant. Look at the Probability of occurrence. Say the t value was computed and probability was above 60%. The t value would happen 62% of the time so we cannot reject the null hypothesis. There is no significant difference between the two means.

### Exercise: qualitative data analysis

As part of a survey to ascertain opinions amongst its clients for falling repeat business, the Cornwell Training Company (which had full residential facilities) conducted a preliminary semi-structured focus group on representatives of some of its largest private and public sector customers. The following is a paraphrased transcript of the group's responses:

Question 1: Having attended at least one training programme at Cornwell, what is your general feeling about the quality of the course and the services offered?

Response:

- Generally the training courses are fine, but there is room for improvement e g the courses could be longer
- It was agreed that shortening the course duration may compromise quality
- Courses of 2 weeks duration were too long for the private sector due to the pressure of work
- The public sector suggested the quality of courses was declining through lack of competent and knowledgeable trainers
- All agreed the courses were generally very dated
- Course fees were too high for some
- General type courses e.g. 'Marketing for non Marketers' were too non-specific, but were better when tailor made to specific industries or clients
- Facilities during the evenings were not very good given the isolated location and few recreational amenities of the Company
- The interior of the building did not match the exterior as beds were small and there were no telephone extensions

Question 2: What do you think could be the reasons for low repeat business to Cornwall?

Response:

- Failure by Cornwell to carry out proper customer training needs analysis
- A Certificate of Achievement was better than receiving a Certificate of - Attendance as was the case now
- Some courses too theoretical and so companies were opting for on the job training
- Acknowledgement of successful acceptance on the course was too long, putting off clients
- Lack of follow-up after the courses

Question 3: In your opinion what can Cornwell do to help customer retention?

Response

- Improve communications and customer service
- Examine or assess the courses as this leads to certification of Achievement rather than Attendance
- Provide better social activities, especially in the evenings
- Have more competent, knowledgeable and interesting presenters
- Better IT facilities

Question 4: Why did you choose to attend a Cornwell course?

Response

- The private sector said it was personal initiative to find suitable courses and then persuading their supervisors to approve it
- Tight training budgets in the private sector dictated cheaper courses, like Cornwell
- The public sector said it was a mixture of the Training Director's and prospective participant's decision
- There was no alternative in the vicinity
- Past reputation

Question 5: Should your organization give you an opportunity to choose any training provider for any course of your choice would you opt for Cornwell?

- The private sector said 'unlikely', unless the issues in response to Question 3 were resolved

- The public sector said they would opt for Cornwell

6

##### Question

Suggest a suitable data analysis technique on the above, carry it out and generate appropriate conclusions

There are a number of ways that the data could be reduced to a manageable format.

Students should examine the Chapter 22 and suggest an appropriate technique including content analysis, narrative analysis and so on.

One simple 'step like content' analysis technique could as follows:

Step 1 arrange the responses into a matrix

Step 2 look for common/different themes or responses

Step 3 develop hypotheses from the common themes or responses

For example:

Step 1

Question

Public sector

Private sector

Both

1. Quality and services

Declining quality

Courses too long

Poor facilities

1. Reasons for low repeat business

X

X

Certification etc

1. How to retain customers

X

X

Better IT facilities etc

1. Why choose Cornwell

Joint decision by Training Director and prospective participant

Personal initiative and persuasion of supervisor

Past reputation etc

1. Free choice of venue

Would opt for Cornwell

Unlikely to opt for Cornwell

X

Step 2

Differences of opinion:

• Divergent to Questions 4 and 5
• Question 1 quality component differences but same issue

Similarities:

• Question 1, 2, and 3 and certain issues in Question 4, e.g. past reputation

Step 3

General conclusions:

• Question 1. Overall agreement on good quality of courses and services; however, private sector concerned about length of courses, fees, and currency of courses. Public sector concerned over declining quality. General agreement on the public sector suggested the quality of courses was declining through lack of competent and knowledgeable trainers
• All agreed the courses were generally very dated, some course fees too high, courses needed better targeting, facilities in the evening were poor, no telephone extensions and beds were too small. Interior did not match exterior image
• Question 2. General agreement on lack of response to course applications, post-course follow-up, courses too theoretical so on the job training growing, lack of Achievement Certification rather than Attendance Certification
• Question 3. General agreement that service quality needs to be good, communications timely, high quality image, provision of Certificate of Achievement, competent, knowledgeable and interesting presenters and better IT facilities
• Question 4. Public sector had a joint Director/prospective participant decision, private sector individual initiative then supervisor persuasion. Reputation, size of training budgets and lack of alternative training providers were other common reasons
• Question 5 Difference of opinion. Public sector would opt for Cornwell but private sector unlikely

From these general conclusions, the researcher may form hypotheses for further testing. Also, these findings could constitute the basis for quantitative research.