Website Worth

Total Pageviews

Thursday

Building links in a Nvivo 2.0 Model

As I have said in another post, Nvivo 2.0 allows us to make links and put the links in to a theoretical/explanatory model. A model is a fascinating feature that Nvivo offers us. But it has been underused for a number of reasons. One reason may relate to ability to build a link.

People are often quite reluctant to explain why a link is established in their study. This 'why' question is, however, difficult to answer. In this post, I will give answer these two questions:

 

1) if I have many people say a same thing and its relation to another same thing, then should I make a link?

2) if I have just few people say a same thing and its relation to another same thing, then should I make a link?

 

People who ask these questions are, actually, confused with the philosophical base of their research. In a qualitative study, researchers tend to use social constructionist perspective rather than an objectivist one.

Let' us review briefly two traditions in philosophy:

- objectivism: the idea that the world is independent of human mind. It is 'objectively' there, whatever you think of it. Therefore, the world could only be measured by scientific methods. (Maybe I should write, in a separate post, how people agree and disagree with each other regarding the world 'scientific').

- subjectivism (with social constructionist perspective as a branch):  the idea that how the world exists depend on human thoughts, perceptions, meanings they give to it. Therefore, the world could only be understood if we ask people to tell us what they think of it.

For instance: We have a public health problem, that is drunkenness. People in the first tradition would look at association between level of alcohol intake and drunkenness (relation between one behavior with one phenomenon). People in the second tradition would be more interested in knowing the association between meanings people give to drunkenness and drunkenness (relation between a certain type of meanings and one phenomenon). People in the former tradition is more interested in knowing frequencies, p-value, confident interval in establishing a link between things. People in the second one, instead, are likely to look at a 'map' of meanings, or the relatedness between things, or the thick descriptions of links between things.

Writers of Nvivo 2.0 are those who on the subjectivism side. Tools within Nvivo, including the modeling tool, are designed to support links between subjective meanings.

But the division between 'subjective' and 'objective' is not rigid, but changeable. That is the key message in this post.

 

In a quantitative data analysis, people must start with finding descriptive and/or inference parameters (mean, media, standard deviation, p-value, r square, so on). But then they have to tell us what these parameters do means. Remember, a lot of people are struggling with understanding and properly interpreting the statistics their fancy softwares (SAS, SPSS)  have just produced to them. Again, that should be the topic of another post. 

Vice versa, in a qualitative data analysis, one must start with finding subjective meanings, constantly comparing the meanings between informants, and putting meanings into 'codes' and 'themes'. When you are sure that a code has same meanings across cases/informants, you can count how many people say a same link. That is to say, after having properly dealt with subjective meanings, you are now interested in frequencies of links between them.

If a link is mentioned by a half of informants or more, it should be put in your model. REMEMBER: Qualitative researchers do not use a sample representing a population like quantitative researchers do. You put it in the model because it represents the meanings (AGAIN, not the INFORMANTS) of that phenomenon.

 

Let us consider the topic 'drunkenness'.

People in quantitative tradition will ask "if drunkenness in my sample represent drunkenness in a reference population?". In this way they have to give a precise definition of 'what is drunkenness' and develop tools to measure drunkenness before they can say anything about prevalence in a sample, and later in the reference population. Representativeness of RESPONDENTS are important for them.

People in qualitative tradition will ask "if meanings of drunkenness in my sample represent meanings of drunkenness in a reference population?". In this way they have to ask people "how do you define drunkenness", "what does drunkenness mean to you?". They will keep asking until the meanings given by informants are 'saturated'. At this point they can stop asking and say the meanings of drunkenness in the sample represent meanings of drunkenness in a reference population. Representativeness of MEANINGS are important for them. Sometimes only few (ten or fifteen) informants can tell you all meanings of drunkenness in a region.

 

Consider this table 1.

 

             Attribute
Themes
Belonging to a minority group Landless Jobless Being male
Drunkenness Informants A, B, C, D, E, F Informants C, D, E, F ....so on... ....so on....
Excessive drinking Informants A, B, C, D .... ...so on... Informants C, D, A, F, E
Making fictive kinship .... ... Informants A, E, F, C  
Swearing brotherhood Informants B, A, D, F      

 

You can see that "Excessive drinking" and "Being male" are 'strongly' related to each other. Then, how we can say that the two themes are 'strongly' related? Because many people mention it.

But again, don't confuse with the number used in a quantitative study and a qualitative study: The number in a quantitative study is frequency of drunkenness as defined by researchers. The number in a qualitative study is frequency of drunkenness as defined by informants. If few experts can define what a scientific definition of drunkenness is, then few lay people can also say what drunkenness means to them.

 

NOW comes the art of a qualitative research: WHAT IF just few people mention a link, what can I do?

If a link is mentioned by less than half of informants, or only few, DON'T throw it away! Now you have to consider a 'refinement' method. The other name for this refinement is triangulation.

Triangulation is an important tool in qualitative data analysis. Its rule is simple: One piece of information is considered valid/true/believable if it is mentioned by different sources of information.

Let's say you have in-depth interviews, observation, and focus group discussions in your study. These methods provide you three different sources of information. If, when in 'the field', you observe one phenomenon, and that phenomenon is mentioned in both group discussions and individual interviews, then you should consider that phenomenon 'believable'.

If only few informants in your individual interviews mention the phenomenon, but you frequently observe it, and it is often mentioned in focus group discussed, you should put it in your model.

 

Here, let's go back to the two original questions:

 

1) if I have many people say a same thing and its relation to another same thing, then should I make a link?

Answer: Yes, but you have to make sure that all people in your study share the same meanings/links. Normally, all informants may share some 'core elements' of the link. It is this 'core' that should be added in the model. When writing your paper, you need to describe all 'variations' around the 'core'. These variations may depend on social demographic characteristics of each informants: being old/young, being female/male, being poor/rich, belonging to a minority ethnic group/majority ethnic group, so on so for.

Tips: In Nvivo 2.0, we can count how many percent of informant say a thing by using "Assay" tool. If you make some variables ('attributes' in Nvivo language) to describe social demographic characteristics of informants, it will be very useful to say about 'variations' in a later stage.

 

2) if I have just few people say a same thing and its relation to another same thing, then should I make a link?

Answer:

- Yes, if the link is repeatedly observed by you, mentioned in most focus group discussions.

- Yes, if the link is not directly observed you (especially some historical events), but mentioned in most focus group discussions.

- Yes, if the link is not directly observed by you, neither is it repeatedly mentioned in focus group discussion, BUT it has the power to link with many other elements in the model, or, it is an important element that help complete the picture you are drawing. Remember: Relatedness is the most obvious feature of a qualitative study.

- REMEMBER: Every piece of information has some meanings. Take your time to discover its meanings before throwing it away :)

Consider table 2. Suppose the link between Being Male and Drunkenness is mentioned by only few informants in your individual interviews.

Being male/ Drunkenness

Directly Observed?

Often mentioned in FGDs?

Does this link help in connecting the other seemingly unrelated links?

Shout it be a link?

Scenario 1

Yes

Yes

Yes

Yes, certainly there is a link.

You witness it.

Scenario 2

No
(it happened long ago!)

Yes

Yes

Yes, it is highly likely to be a link. More than one sources of information mention it and it has power to link.

Scenario 3

No
(It occurred in the past or you just could not see it)

No

Yes

  1. YES, if you decide only few informants have made the link because informants ARE constrained by something in answering your questions (sensitivity of the issue, the interview settings, and the 'taken for granted' beliefs. See table 3)
  2. NO, if you decide that the informants are NOT constrained by anything (as mentioned above) in answering your questions. REREAD the informants' transcripts to make sure the link are made by INFORMANTS, not YOU.

Scenario 4

No

(It occurred in the past or you just could not see it)

Yes

No

YES, it is likely to be a link. The link are mentioned by more than one sources of information. But ALSO consider factors in Table 3.

Scenario 5

Yes

No

Yes

YES, it is likely to be a link.

You witness it.

Scenario 6

No

No

No

NO. But consider factors in Table 3 if it could be "YES".

Scenario 7

Yes

No

No

YES, you witness it.

Scenario 8

Yes

Yes

No

YES, you witness it, and the link is mentioned by more than one source of information.

 

Table 3. Informants could be constrained in answering your questions in following ways.

Constraining factors Ask yourself
Sensitivity of the topic Example: Is the link less mentioned because I am asking them about their sexual practice?
Interview setting Example: Is the link less mentioned because I am a 'stranger'? Or it is because of the presence of somebody else ?
Taken for granted belief Example: Is the link less mentioned because the informants hold a certain important belief but they think it is unnecessary to speak out?
For instance, many men say " I am drunk because I 've just visited a Muong village". Here they make a link between "being drunk" and "visiting a Muong village". The link that is hidden is: "being drunk" and "drinking norms in Muong village". The men take it for granted that going to a Muong village is to be drunk. Then, the research should consider putting 'drinking norms in Muong village' into his/her model, even though this is mentioned by only few informants.

I enjoy writing this post.