6.1 Questionnaire
6.1 Questionnaire
6.1.1 What is a questionnaire?
The questionnaire is one of the most commonly used data collection instruments at the disposal of the researcher (Blaxter et al.,1996;Bryman,2008;Miller and Brewer,2003;Wilson and McClean,1994).This is because the questionnaire survey has many advantages in social research in terms of amount of data it is possible to collect,cost,time,involves greater anonymity,and considerable accuracy (Denscombe,2003;Kumar,2005).Questionnaire surveys are a useful method by which information can be obtained in a structured format and can be administered without the direct support of a researcher.Its design process assists in identifying the associated analysis stages (Wilson and McClean,1994).
The questionnaire is superior to the interview method for the purposes of largescale surveys.Within a relatively short period of time in comparison to interviews,the researcher can send out many copies of a questionnaire to a large number of respondents.The questionnaire may therefore be seen as providing a potentially unlimited number of responses.
With regard to cost,a questionnaire survey can be relatively cheaper than research through interviews in terms of time (Denscombe,2003).For example,a questionnaire to be completed by hand will only involve the expenses of printing,paper,and postage.Such surveys require relatively little of the researcher's time if respondents complete the questionnaires themselves.
The questionnaire survey requires respondents to answer a standard set of questions in a fi xed format.Such surveys are very highly structured (Wilson and McClean,1994) and benefit the researcher by obtaining precise and comparable data.Some argue that the problem of bias is reduced in a questionnaire survey because such surveys allow a large number of responses to be collected.The fact that the questionnaire survey does not involve face-to-face contact between the respondent and researcher may eliminate the psychological barrier which can arise between individuals who do not know each other and also provides greater anonymity (Kumar,2005).Anonymity may help to increase the likelihood of obtaining accurate information when sensitive questions are listed (Kumar,2005).
Although questionnaire surveys have several advantages,it is important to note that they also have disadvantages.Such disadvantages are reflected in aspects of design,limitation of questioning scope and limitation of flexibility in terms of response format (Wilson and McClean,1994).In respects to the two limitations,we can infer that researchers draft the questions of the questionnaire according to their study aims.This design may therefore reflect the subjective opinions of the researchers and the questions may not provide the respondents opportunity to respond as fully as they would wish.Moreover,such questions do not allow the respondents the opportunity to answer in their own words because of the closed structure of the questionnaire.In some instances,the respondent may not understand the questions and is therefore unable to respond fully,and it may be difficult to fully explore the specific experiences,feelings and attitudes of the respondents.Researchers will derive mainly factual information from such surveys.Frequently,the copies of the questionnaire are not all returned.Therefore,the problem of missing questionnaires will exist in a questionnaire survey.
6.1.2 Why a questionnaire?
I used questionnaires in my research because of the advantages as a research instrument this method has.A particular advantage is that a questionnaire survey allows a large amount of data to be collected in a relatively short time.My own experience of using a research questionnaire is a good example as I collected 232 samples of questionnaire within two months.It would have been impossible to conduct such a large number of interviews in such a short space of time.The more questionnaire data researchers collect the more information they will be able to obtain (Denscombe,2003).I therefore concluded that a questionnaire would be beneficial in my research.
The use of a questionnaire was also appropriate to my research aims.As emphasised by some scholars,research techniques do not exist in a vacuum(Bryman,2008).They are tools,which connect closely to the practical research.The aim of my research was to investigate what forms of psychological violence are used by perpetrators.Psychological abuse may take both verbal and non-verbal forms.In respect to the former,this may include ridicule,swearing,cursing,dirty language,etc.But with regard to the latter,this may encompass facial expressions,threatening gesture,non-communication,etc.I also wanted to investigate how much psychological violence occurs between wives and husbands,whether the level varies between intellectual and non-intellectual families,how victims experience such violence,and how such violence impacts on victims (usually wives).Using a structured questionnaire survey,I could design detailed questions from which I could obtain a good volume of factual information in a relatively limited time.
For these reasons,I chose to use the 'questionnaire' instrument to conduct part of my research.I next had to consider the design of the questionnaire.The following section describes the design considerations and process.
6.1.3 Questionnaire design
As Warwick and Lininger (1975) point out,questionnaire design has two basic goals.The first is to obtain information relevant to the purposes of the survey and the second to collect this information with maximum reliability and validity.Good questionnaire design may therefore play a decisive role in social research.Every researcher needs to think about this carefully.Questionnaire design includes two aspects:the external and internal.Neither should be neglected.Accordingly,let us look at how I approached the design of my questionnaire.
6.1.3.1 The external design
As Wilson and McClean suggest (1994),researchers need to think of any possible cultural differences when they design questionnaires.Therefore,based on my knowledge of the Chinese context,there was no title on my questionnaire.Domestic violence,as I pointed out in Chapter 1,is a new concept and term in China,and it is a very sensitive topic.If this term had been used in the title of the questionnaire,it may have influenced respondents' willingness to answer it.Although Chinese people are gradually realising that domestic violence is not a private matter and should be stopped,they still may not like completing a questionnaire filled in this way.I therefore used a non-stimulation style in my questionnaire design.As I needed to ensure the cooperation of as many respondents as possible,I dispensed with a title.
In respect to the external aspects,clear presentation of the questionnaire is important.Self-completion questionnaires,as Dillman (1983) said,should not be cramped in terms of layout.An attractive layout is likely to enhance response rates.It is easy on the eye and facilitates the answering of all questions.Most questions in a self-completion questionnaire are likely to be of a closed kind.Therefore,researchers should consider whether the answers should be arranged vertically or horizontally.Usually,the length of answers will dictate a vertical arrangement.Questions and answers should be kept together in the questionnaire.Respondents should be able to read and answer the questionnaires easily.This prevents the respondents from forgetting to answer a question or providing an answer in the wrong space.The layout of questionnaire should make it clear how the researcher wants the respondents to complete their answers.The better the questionnaire design,the more information the researcher will obtain.
According to the principles of external questionnaire design,I paid careful attention to layout.For example,I arranged the fixed answers vertically and horizontally.Its line space was 1.5 times.Questionnaires were printed on one side,only of three pages,and on good quality paper.The questionnaire was written in Mandarin.
The 'introduction' to the questionnaire,like the layout and question order,is also important in determining the respondents' feelings towards the questionnaire(Wilson and McClean,1994).In my introductory remarks,I stated the purpose of this questionnaire and request as well as thanks for respondents as follows:
[Family] is a cell of our society.It plays a great role in the society.It is said that harmony in the family will benefit everything.Therefore,we should pay more attention to family life.Raising the quality of family life is our duty.However,how do we improve it? It is hard work that needs to be researched.Please will you complete the questionnaire in order to create a beautiful life in the future? Thank you very much.
I introduced clear instructions in each section,guiding the respondents to answer the questions.In accordance with the principles of good questionnaire design,I kept questions and answers together.I also ordered the questions logically,grouping similar questions together.I thought that it would be easier for respondents to follow the questionnaire by structuring it in this way.
In addition,I paid attention to the psychology of respondents when designing the external aspect of the questionnaire.The number of questions needed to be appropriate,too many might influence the quality of the respondent's response.Researchers need to consider the nature of the subjects/respondents and design the questionnaire accordingly.For example,my subjects/respondents were married people,not young people or students,and their attitude towards the completion of questionnaire might be different from that of young people.Some of them might complete it carefully but others might do so randomly and the risk of this might increase of the questionnaire was too long.Therefore,I designed 37 singlechoice questions requiring a simple 'Yes' or 'No' answer in the main body of the questionnaire.
6.1.3.2 The internal design
With regard to the internal design,the content of the questionnaire is a key to the quantitative approach.The design must reflect the research intentions and aims clearly.The design process can be seen as the translation of the broad objectives of the study into questions (Warwick and Lininger,1975).The research needs to have a clear hypothesis,which will then determine which variables need to be included.Researchers need to consider whether these variables are dependent or independent,and to intervene or control them,and to anticipate the type of analysis that may be conducted.
In my research,I have a clear study aim set out in my introduction to Chapter 5.My research objectives are to reveal people's behaviour (in relation to married women and men),their experiences covering their emotions and their attitudes to domestic violence,particularly to psychological violence;to explore the nature of psychological violence in China,including its extent,forms and impacts on the victims;and to explain the reasons for it occurring between wives and husbands.My questionnaire would therefore focus on the marriage,the family,and domestic violence or psychological violence including the behaviour and attitudes of those experiences involved and impact on the emotions of victims.
Various types of questionnaire design are available to the research.There are two main types:the closed and the open format,which may be applied in the design of both structured and unstructured questionnaires.In respect to a closed or structured question,respondents are limited to a set of fi xed statements from which they have to choose the most appropriate answer.However,open or unstructured questions allow the respondents to reply more freely.It is from open or unstructured questions that researchers may obtain more 'precise' and subtle original data.We can see that these two types of questions achieve very different things.The former will constrain the respondent's reply and may make something difficult for them,while the latter will be more time-consuming and result in a more difficult coding exercise for the researcher.
In practice,which type of such design is used by a researcher depends on study ideas and aims of the researcher.Thus,I decided to use closed questions in my design.This was because I wanted to obtain factual information,which closed questions would generate as set out in Section 6.1.2.This also reduced the amount of time respondents would need to spend completing the questionnaire and me in coding.The questionnaire design was therefore very structured and standardised so respondents could answer the questions easily.Additionally,the raw coding could be outlined by the researcher during the design process.I also decided to use an indepth interview,an open format in my research,and this interview would allow me to get 'precise' information which I would not be able to obtain from a selfcompletion questionnaire.This will be a complementary exercise.The in-depth interview will be discussed in a later section.
My questionnaire was divided into two sections.The first section was concerned with demographic information and invited the respondents to provide personal information about themselves and their spouses.The questions cover aspects such as age,sex,place of birth,place or area of residence,level of education,occupation,marital or divorced status,marriage or divorced age,income and financial situation,office rank,academic and technical titles,and hobbies and interests.Knowing the'sex' of the respondents helped me to explore the gender issues.The information on'educational background','occupation' and 'academic and technical titles' was very relevant to my investigation of whether domestic violence,particularly psychological violence,manifests itself differently between intellectual and non-intellectual families.
The second section formed the main body of my questionnaire.The questions focused on specific family problems such as a conflict between partners,behaviour,emotions and impacts caused by physical or psychological violence,and attitude towards marriage.
There were 55 questions in this section dealing with psychological violence as a form of domestic violence.Twenty questions dealt directly with verbal abuse and two with verbal abuse indirectly,while eight questions dealt with non-verbal abuse.Six questions were concerned with the nature of the psychological violence.Eight questions plus four questions were connected with emotions and symptoms of ill-health caused by verbal and non-verbal abuse respectively.One question was intended to identify what factors,including psychological factors,which caused domestic violence.Three questions asked about attitudes to marriage and family,while two dealt with aspects of behaviours.The final question dealt with the phenomenon of domestic violence in Chinese families (see Appendix 1).
Thus,it can be seen that the content of these questions linked closely with my research aims and were designed to attain my study objectives.
6.1.4 Questionnaire sampling
Before talking about my data collection from the questionnaire,I need to have a brief look at the literature on what sampling is and its kinds and features.Sampling is a useful short-cut and may be leading to accurate results (Gorard,2003).Basically,sampling is divided into two kinds by social researchers (Bryman,2008;Denscombe,2003).The first is known as 'probability' sampling,while the second as 'nonprobability' sampling.
Usually,probability sampling is seen as good practice.This kind of sampling is based on the idea that the people or events that are chosen as the sample are chosen.The researcher may have some notion of the probability that these will be a representative cross-section of people or events in the whole population being studied.Within this probable context,a sample is selected by random selection so that a unit in the population has a known chance of being selected (Bryman,2008).
However,many researches are based on non-probability samples (Bryman,2008).This sampling is conducted without knowledge about whether those in the sample can be representative of the overall population (Denscombe,2003).Usually,nonprobability sampling is used because the researcher may find it impossible to obtain probability samples.This sampling may be seen as an umbrella because it captures all forms of sampling that are not conducted in the light of the canons of probability sampling (Bryman,2008).It covers a wide range of different types of sampling strategies such as convenience sample,the snowball sample,the quota sample,and so on (Bryman,2008).
Based on the sampling principle,I chose non-probability sampling in my selecting the sample.This is because according to my research,I felt it was not feasible to include a sufficiently large number of examples in this study.This is also because I did not have sufficient information about the population to undertake probability sampling.I only wanted to explore how the respondents in my survey used/ experienced domestic violence in the home.In addition,my sample may be seen as convenience sampling because I collected my questionnaire data accessibly,which will be discussed in the next sections.Convenience sampling is built upon selections that suit the convenience of the researcher,which can be called 'first to hand'(Denscombe,2003).Through sending questionnaire to the Danwei or the respondents,I received the questionnaires back within the limited time and cost available.
As a result,I obtained interesting findings from these questionnaires.This has been reflected in the next chapters.Nevertheless,there is a problem occurring in such a sampling strategy because the researcher cannot generalise these findings.We do not know of what population the samples are representative.For example,my questionnaire samples were simply a group of individuals and couples who were available to my research aim but they could not be representative of couples as whole.But noticeably,this strategy is used commonly and is acceptable in social research.As Bryman (2008) suggests,convenience sampling may play a more prominent role than is sometimes supposed.It is different from probability sampling which always involved a lot of preparation.Convenience sampling represents too good an opportunity to miss,especially when opening up a new research area.
6.1.5 Questionnaire samples
As Bryman (2008) emphasises,a research design provides a framework for the collection and analysis of data.In this section,I will describe how I used my questionnaire to collect my data.I will also comment on my experience of collecting the questionnaire data.
Through using personal contacts for developing networks,I was able to collect 232 questionnaires completed largely between September and October 2003 in China.Before this collection,I had a basic plan about who I would contact and ask to help me.These were some of my classmates,friends,and even my relatives.For example,Ms Lu[1] and Mr Li[2] were my university classmates.With their help,82 respondents[3] were found from universities.In addition,during my data collection,I met these people who were heads of 'Danwei' (work unit) and explained my research aims to them and described the kind of respondents I needed.After my detailed introduction and explanation,they then supported me and distributed my questionnaire to their staff who were married.Because of their support,I was able to collect my data successfully within a very limited time.
The questionnaire data was collected in three different cities in the Hubei Province of South-Central China.Such collection in a wider range of regions may enhance the validity of the data.The three cities had different characteristics.Wuhan,for example,is a large city,the capital of Hubei Province.Its population is about seven million,of which there are three million in the city centre.Wuhan is also my hometown,so I could collect data there very readily.Jingzhou is a city of light industries with an historic past,with a population of about six millions,including half a million population in the city centre.This was in fact the first city where I collected data initially.Xiaogan is a new developing city,near Wuhan.Its population is about five million,including 300,000 populations in the city centre.I collected data there because it is the city,which has the only women's refuge in Hubei Province (the Hubei Xiaogan Protection Centre for Women and Children).Its work against domestic violence has often been praised by the Hubei Women's Federation.I was fortunate enough to get excellent support from the Xiaogan Women's Federation.
The samples were collected from subjects in different occupations including teachers,doctors,cadres,policemen,engineers and other workers,which might be representative of both intellectual and non-intellectual groups.My initial plan had been to collect data only in universities and academic institutions.In practice,however,I found that things developed differently and needed to be amended.As a result,my data covered a wider range and scope of jobs and the respondents had a wider range of educational backgrounds than originally planned.I therefore had an opportunity to establish whether they behaved differently in the home and had different attitudes to their marriages and families.This accords with my research aim and objective of establishing whether psychological violence occurs in both intellectual and non-intellectual families.
During this exercise,I firstly left the questionnaires with contacts who circulated them to their staff for completion then after a period of time,I returned and collected the completed questionnaires.As a result,the 232 samples were collected voluntarily in different 'Danwei'.In addition,I had a number of ideas in order to help me to get the maximum number of questionnaires back.For example,I asked the people who helped me to send out the questionnaires to have ready a large envelope for the collection of the returned questionnaires.In order to eliminate respondents' concern about confidentially,I felt this helped my getting back the questionnaires smoothly.At the same time,by using this means of data collection,I did not have close control of its collection.For example,there was imbalance between females(n=128) and males (n=104).This is illustrated in the tables below.Apparently,my collection data had truthfulness but not artificial control.
6.1.5.1 Distribution
Of the 270 copies sent out,232 copies were returned,128 by females and 104 by males,all of whom were married.This represents an overall response rate of 85.9 per cent,which is very high.The respondents were from three cities in Hubei Province:Wuhan,Jinzhou and Xiaogan[4].The questionnaires were collected from several different 'Danwei' [5].The distribution of my samples is shown in the tables below.
Table 6.1.5 (a):232 respondents from three Cities[6]
Table 6.1.5 (b):232 respondents from the different Danwei[7]
The distribution of the samples is shown in Tables 6.1.5 (a) and (b).From them,we can see that 156 respondents (67.2%) were from Wuhan,and 51 (22%) and 25 (10.8%) from Jingzhou and Xiaogan respectively.There were 128 female respondents (55.2%) and 104 male respondents (44.8%),and the samples were collected in seven Danwei.Thirty-five per cent of the samples were collected in universities.From the distribution,we can infer that my samples had a wide range of occupations.This is in accordance with my research aims.
6.1.6 Reflection on questionnaire method
From using the questionnaire in practice,its merits and weaknesses were exposed,and I learned lessons for my future research.I will mainly deal with how the respondents answered the questionnaire in this section.This could be seen as an assessment of my questionnaire design.From it,we will see whether the design was good or not.It is a key issue for every researcher.Two possible problems of assessment were highlighted in feedback from my respondents and myself.
Two very different views of the content of my questionnaire were expressed by the respondents.One was that it was very appropriate.However,the other was that my questions were naïve.The respondents holding the first view were largely female;they felt that the situations reflected in the questionnaire were representative of relationships in their families.The respondents holding the second view were mainly males.They commented that they felt the designer might be unmarried and without experience of marriage life.These male respondents thought that behaviour such as using dirty language,non-communication with their wives,and so on were the norm in couples,and should not be seen as domestic violence and were not worth researching.Interestingly,we can see that gender issue affects feelings about my questionnaire design.
Upon reflection,I feel that the main body of the questionnaire (Q[8]1-55) was focused on my research aims.The majority of the questions covered the level of psychology I explored.However,the weak features of the questionnaire were revealed after the data was analysed by SPSS[9].For example,some respondents did not answer all or some of the questions;missing data immediately causes problems for the researcher.Let us look at the incidence of this problem.
In Part I of the questionnaire,the respondents answered the personal and factual questions (Section A),more fully than the questions relating to his/her spouse(Section B).There were 16 questions dealing with the respondent's personal information,but there are eight concerning their spouse's personal information.Table 6.1.6 (a) below sets out.
Table 6.1.6 (a):Missing situation of A (Q1-16) and B (Q1-8):
Table 6.1.6 (a) suggests that the design of Section A of Part I of the Questionnaire was better than the design of Section B.For Section A,the proportion of missing answers was lower.Of the 16 questions,the proportions of those not answering 10 were below one per cent (0-0.8%),which was 62.5 per cent of the respondents answers.The proportion of those answering six questions was over 11 per cent,which reached 37.5 per cent completion rate.However,for Section B,this situation was worse.The rates of missing answers were very high.Of the eight questions,the proportions of answers to each half (four questions for each) were under 10 per cent and just over 20 per cent respectively.From these results,we may infer that the respondents preferred or found it easier to answer questions about themselves.
In comparing the completion rates for Sections A and B,we can see that as Denscombe (2003) states,it is not good practice to ask respondents to answer for other people although some matters of fact may be included,provided they are relatively easy for the respondents to answer.From this practice,I think that it might have been better if I had not designed or included 'Section B'.The questionnaire would have been shorter so the respondents might have been able to spend more time on answering the other questions.I now feel that Section B was superfluous and that it perhaps caused respondents to be less cooperative (Wilson and McClean,1994).
Regarding Part II of the questionnaire (the main body),the missing answers largely occurred in the questions offering multiple-choice responses.Of the 55 questions,40 required a single-choice response (72.8%) and 15 a multiple-choice (27.2%).First,let us look at the overall number of missing responses in Part II of the questionnaire then compare the response rates to questionnaires with multiple-choice responses and single-choice responses.This is set out in the tables below.
Table 6.1.6 (b):Overall missing answers situation in Questionnaire II (Q1-55):
Table 6.1.6 (b) suggests that the design of Part II of the questionnaire was good at getting people to answer.Of the 55 questions,32 had a very low missing proportion of below 10 per cent (0-10%).This was over half percentage (58.2%).However,23(41.8%) had a quite high missing proportion (11.2-29.3%).There were particularly 12 had a very high missing proportion that was from 20.7 per cent to 29.3 per cent.How did the missing situation distribute between single-choices and multiplechoices? The table below compares the missing proportion between the single choice and multiple-choice response in questionnaires.
Table 6.1.6 (c):Comparison of missing situation between single-choice and multiplechoice responses[10]:
Table 6.1.6 (c) clearly shows that there were substantially more missing responses to the multiple-choice questions.Of the 15 multiple-choice questions,no question was at 'zero' missing answer level.Two had a missing proportion of below 10 per cent.Another two had a missing proportion between 11.6 and 15.5 per cent.Eleven had a very high missing proportion of over 18 per cent.Of the 55 questions,the missing proportion of these 11 multiple-choice responses occupied 20 per cent.
However,of the 40 single-choice questions,30 had a low missing proportion from zero to 10.3 per cent.Such proportion was 54.6 per cent among 55 questions.Eight had a missing proportion was between 11.2 per cent and 17.7 per cent.But two had a high missing proportion between 27.6 per cent and 29.3 per cent.
This suggests that questions with a single-choice of answer are more likely to be answered and are therefore better than multiple-choice questions.Why should this be? I infer that the respondents found the multiple-choice questions difficult to answer in a short time or were bored by them.I also infer that the respondents might not understand the reason for asking such questions so they did not answer them.There may have been two reasons for this.One,because my instructions for answering the multiple-choice questions might not have been sufficiently clear.Two,it might be because I did not hand out the questionnaire personally to the respondents.As a result,I could not explain the purpose of the questionnaire fully to the respondents,which may have influenced both the number of missing answers to the multiple-choice questions and to the single-choice questions.The questionnaire as a quantitative research tool in social research thus has its limitation.
However,having reviewed the number of the missing answers,it does seem that overall design of the questionnaire was good.There are relatively few missing answers reflected in Section A of Part I of the questionnaire and in the single-choice responses in Part II of the questionnaire.From these results,we can conclude that the respondents were happy to provide the personal and factual information and preferred the single-choice to the multiple-choice questions.In the light of my experience,I have come to understand that in designing questions or choices for respondents they should be kept as simple,direct,familiar,clear and specific as possible (Warwick and Lininger,1975).
6.1.7 Data analysis
Data analysis is an essential step in the process of research because researchers need to make their data speak (Ramazanoğlu et al.,2002) in order to identify and explain the issues.However,what is analysis in particular? What is quantitative analysis and how did I analyse my questionnaire data?
6.1.7.1 What is analysis?
As Bernard (2002) stresses,analysis is a search for patterns in data and for ideas that help researchers explain why these patterns occur.Denscombe (2003) specifically points out that analysis involves the separation of things into their component parts,that is,it involves the dissection of complicated things in order to identify their basic elements.From Bernard (2002) and Denscombe's (2003) definitions,we can see that researchers,through analysing data,will discover the key components or general principles underlying a particular phenomenon,which can then be used to provide a clearer understanding of it.
In respect to quantitative analysis,this is linked to numbers and can be called statistics (Punch,1998).For example,it is suggested that quantitative analysis is based on systematic,statistical relations between numbers (Alasuutari,1995).At the starting point of such analysis,the data is transformed into a table format and the observation units are given values in different variables.Additionally,argumentation is based on average relations and as such the starting point is to search for differences between observation units in terms of different variables(Alasuutari,1995).The principle of such analysis is to find statistical regularities in the way different variables are associated with each other.
Most methods of quantitative analysis,such as factor analysis,cluster analysis,regression analysis,etc.,are really techniques for processing data and for finding patterns in it (Bernard,2002).Therefore,such analysis normally follows a series of steps in order to get the information from the quantitative (raw) data:quantitative raw data→editing→coding→analysis (Kumar,2005).
Faced with a large number of questionnaires (raw data),how do researchers analyse,including edit,code them by counting them manually or by using a computer? For the sake of accuracy and speed in data analysis,SPSS is a widely used and comprehensive statistical social science research programs (Bryman et al.,1999).The great advantage of using a package like SPSS is readily apparent;it enables researchers to score and to analyse quantitative data very quickly and in a variety of ways.This saves the researcher's time and eliminates mistakes in working out scores and carrying out complex calculations.Using SPSS,a large volume of questionnaires or complex data can be changed into readily understood clear data.For example,the responses to the questions in my questionnaire such as 'yes','sometimes','no','much','more','less','little',etc.,or letters of alphabets were metamorphosed into numbers in tables within only two or three months,and provided valuable information for my research.How I went about analysing my questionnaire data and how my analysis progressed in practice is described below.
6.1.7.2 Analysis of data
As Miller and Brewer (2003) state,measurement is not only very important in the quantitative approach but it has to be as exact as possible.This is because the quantitative approach is a very structured approach involving the numerical measurement of specific aspects of phenomena.Explanation of measurement must be formulated in terms of the relationship between variables.I took care to check my data thoroughly and tried to minimise errors occurring in both the data and the variables.Specifically,I often examined the typing answers by respondents and amended them when I found my inputting errors,which guarantees the correct data.
Before computer analysis,the researcher should devise a codebook that provides a set of rules for assigning numerical values to the data obtained from the respondents(Kumar,2005).This prepares the data for inputting into SPSS.For example,in respect to 'Place of Residence',I coded 'Wuhan=1','Jingzhou=2','Xiaogan=3',and'Missing=99'.Through doing the codebook,I understood that the researcher needs to read these original materials (questionnaire) and carefully to create the codebook.The researcher needs to amend their codebook as soon as possible when they find any missing information.For example,when I coded 'Native place or area',I missed a few provinces because I was looking at so many questionnaires (232) then I quickly supplemented them into my codebook when I discovered this.Therefore,this step can be seen as one needing care.
Researchers need to name as many 'variables' as possible,so that they can easily check their data and this also allows them a wide range of options when editing/ coding the collected data through structure inquiries.For example,initially I did not give a unique identification (ID) to every respondent.This meant that having analysed the data using SPSS,I could not identify the respondents in the analysis with the respondent's original questionnaire.I thereafter defined ID for every respondents and I no longer had any difficulty identifying the respondent no matter how often I analysed the data using SPSS.
In addition,I named 'variables' as specifically as possible,which was useful for both my present and future research.For example,with regard to the 'Age' of the respondent,there were two variables:the exact age and the age group.I did the same with 'Age on Marriage'.From my analysed data,I could therefore see this age group and at which age domestic violence occurs more frequently.The more specific or detailed the variables,the stronger and more powerful the analysis will be.Overall,there were 179 variables in the questionnaire,arising from 24 questions in Part I and 55 questions in Part II.
I adopted a rigorous attitude towards the analysis of my questionnaire data because errors must be kept to a minimum in research.Social researchers,through the investigation and exploration of social phenomena,seek explanations of the related phenomena and issues.This will be jeopardised by errors in data.When coding the raw questionnaire data,I typed in the data one question at a time.After coding,I structured the data thoroughly and immediately corrected any errors I discovered.Ideally there should be no errors in the data.I improved and supplemented my coding as this work progressed and I had new ideas.This is a very important aspect of researcher because the process of research is the one of discovery.Through this analysis of my raw data,I came to understand that quantitative data analysis requires hard,complex and careful work.
Having finished coding the questionnaire data,researchers need to develop a framework of analysis for their quantitative studies (Kumar,2005).Within such a framework,they need to consider the variables and how to anlayse them.They have to decide the crosstabulations that need to be carried out and which variables need to be combined to construct the researcher's major concepts,and also to decide which variables are to be subjected to which statistical procedures.Such decisions help researchers to make further sense of the data.For example,through cross-tabulation,researchers may discover a relationship between an event and subjects or a gender difference.Therefore,this step in analytical process directly fuels discussion and conclusions.
In line with the research aims described Chapter 5,I firstly chose to analyse frequency distributions from the variables.This was because I needed to understand how much psychological violence occurs in Chinese families;how frequently couples use verbal or non-verbal abusive behaviour in the home,and how respondents experiencing domestic violence feel etc.Secondly,I chose to analyse cross-tabulations from sets of variables concerning the relationship between wives and husbands.All the cross-tabulation tables,graphs and pie charts in my thesis reflect the relationship between the issues of psychological violence and gender or educational background.This is because my research aims require that I explain whether there is a difference in psychological violence occurring between wives and husbands;and whether the incidence of psychological violence in intellectual families is greater than in nonintellectual families.I therefore chose to examine frequencies and cross-tabulations to test for significance especially with regard to gender and level of educational achievement.The results are set out in Part III (Research Findings and Discussion).