From theory to data
|Faculté||Faculté des sciences de la société|
|Département||Département de science politique et relations internationales|
|Cours||Introduction to the methods of political science|
When we do research and in particular quantitative research, that is, research that is part of the post-positive paradigm, operationalization is the key moment in the process. Without a good operationalization, one cannot make a relevant research, because it is a formalized structure.
- 1 Scientific Research
- 2 Theory and hypotheses
- 3 Operationalization
- 3.1 Definition of operationalization
- 3.2 Phases of operationalization
- 3.3 Unités d’analyse
- 3.4 Criteria for distinguishing variables
- 3.5 Types of variables
- 3.6 Relationship between concepts and indicators
- 3.7 What is the relationship between the concepts and the indicator?
- 3.8 Empirical translation of complex concepts: phases of the operationalization of complex concepts
- 3.9 Empirical translation of complex concepts: examples
- 3.10 Errors in the transition from concepts to indicators
- 3.11 Reliability and validity
- 4 References
Scientific Research[modifier | modifier le wikicode]
How do you do a search?[modifier | modifier le wikicode]
According to Corbetta, the definition of scientific research is a creative process of discovery that develops along a predetermined itinerary and according to pre-established procedures that have been consolidated within the scientific community.
First of all, there is the idea of a creative process, because research is also about creation, we seek to discover something. What is important is to follow specific procedures that are pre-established and consolidated within the scientific community.
Empirical research[modifier | modifier le wikicode]
According to Raymond Boudon, "quantitative surveys are those that collect comparable information on a set of elements from one element to another. It is this compatibility of information that then allows the enumeration and more generally the quantitative analysis of the data..
Empirical research must develop within a framework that is collectively shared. It is a process where research is collective, because it is based on a process produced by others, the process must also be public with the idea of transparency that is important in research. Everything must be transparent, controllable by others. All the procedures implemented must be controllable by others with the idea of replicating what has been done, everything must be replicable. Research is a collective and public process that must be subject to criteria of transparency and control.
Another criterion is that of cumulability, Newton said: "If I have seen further than others, it is because I have been carried by the shoulders of giants". The researcher can make a discovery because he or she can rely on the research of other researchers.
The researchers' collective frame of reference is structured around two moments:
- in the logical structure of the research process.
- in the choice of technical instruments used.
It is at these two moments that the collective research frame of reference is seen, expressed and found.
The five phases of the process and the research question[modifier | modifier le wikicode]
Corbetta makes a difference between the phases of research and the processes that allow a better understanding of the different stages of these processes.
1) Research Question
2) Theory: formulating a theory or relying on a theory.
→ deduction (from the general to the particular): hypotheses are deduced from a theory, from the general to the most specific.
3) Hypothesis: hypotheses stem from an existing theory, we will try to verify them, falsify them through research.
→ operationalization phase: construction phase of the research drawing.
4) Data collection: is the collection to empirically test the hypotheses. This concerns the work plan, there are a number of decisions to be made such as the type of data, the number of cases to be analysed, the location of cases, how to select them and the method of collection.
→ data organization: distinction between information and data, the data have been organized; we will create a data matrix according to the quantitative approach. Data is the raw material that must be organized in order to dismantle or test a hypothesis; it is information organized in such a way that it can be analyzed.
5) Data Analysis
→ induction: we go up to the generality and we come back on the theory; connected by a method of feedback to the theory passing from the particular to the general through the results. There is the idea that the results will be used to create theories and analyze hypotheses.
In the reality of research, the steps are often distributed in a different way; often assumptions are developed after the data have been collected. Sometimes the theory is developed after having analysed the data, sometimes during the empirical phase, sometimes the theme is new and unknown that is why a purely descriptive research is made, sometimes the data collections do not start from a specific theory, because we want to include a wider field that allows to analyze several hypotheses.
Process of operationalization of concepts[modifier | modifier le wikicode]
There are two phases to translate theoretical concepts into something empirical.
- operationalization of concepts: transforming concepts into variables, variables being something that can be manipulated while concepts cannot be processed because they are abstract.
- selection of research instruments: data collection instruments and procedures.
Theory and hypotheses[modifier | modifier le wikicode]
It is the process of "deduction" that makes the link between theory and hypothesis, therefore derives from theory. However, it is difficult to distinguish between theory and hypotheses.
Theory[modifier | modifier le wikicode]
According to Corbetta, a theory is a set of propositions that are organically linked to each other and that is at a high degree of abstraction and generalization with respect to empirical reality, which derives from empirical regularities and from which empirical predictions can be made.
- set of proposals: this is not one proposal, but several proposals, they are articulated and interrelated.
- abstraction of proposals and generalization: the theory is at an abstract level. A theory is something that is intended to be general.
- Theory derives from empirical regularities: the idea that theory comes from previous research and from empirical regularities that have been observed systematically and can be found in different contexts.
- allows empirical forecasts: allows forecasts to be made according to conditions and context.
Hypotheses[modifier | modifier le wikicode]
According to Corbetta, a hypothesis is a proposition that implies a relationship between two or more concepts that are at a lower level of abstraction and generality than theory and that allows the theory to be translated into empirically controllable terms.
- level of abstraction and generality lower than theory: assumptions are specific.
- provisional nature of the hypothesis: hypotheses are subject to control by being verified and falsified, a hypothesis is never definitive.
Difference between theory and hypothesis[modifier | modifier le wikicode]
The essential difference between theory and hypothesis is that theory and a more general and abstract set of propositions whereas hypotheses are not specific enough to be variables, they are theoretical concepts, but a little less abstract.
The difficulty lies in the fact that we are in the gradation, one is a little less abstract than the other. The hypothesis allows us to go into the field in a direct way.
A theory must be able to be articulated into one or more empirically controllable hypotheses that can be transformed into a series of hypotheses. This is the criterion of scientificity, the theory combines theoretical propositions.
Criteria for the Scientificity of Assumptions[modifier | modifier le wikicode]
There are three important elements:
- the assumption should not be too general.
- a hypothesis must be positive in the sense that it must not include a normative dimension, there must be no judgment.
- a hypothesis must be formulated in such a way that it is falsifiable.
The controllability of a hypothesis is fundamental, we must be able to compare it with field data. The hypothesis must be controllable.
According to Popper and Kuhn, a hypothesis must be falsifiable. This gives the hypothesis a scientific character, because a good hypothesis must be refutable. For example, according to Popper, the proposition "god exists" is not a hypothesis because it is not refutable; however, the hypothesis "all swans are white" is falsifiable because they may be of a different color. Thus the main characteristic that gives the relevance of a hypothesis is the fact that it is falsifiable.
- falsifiable hypotheses: heavy objects tend down if nothing holds them back; it rains every Thursday.
- non-falsifiable hypotheses: either it rains or it does not rain; all the points of the circle are equidistant from the centre.
To sum up, the hypothesis to meet the criterion of scientificity must be falsifiable.
Examples of hypotheses in political science[modifier | modifier le wikicode]
Value Change Theory[modifier | modifier le wikicode]
Value change occurs through the replacement of successive generations of people. It is postulated that in post-war Europe there has been a transformation of value systems, from materialistic - security of material value, physical security - to post-materialistic - values linked to personal fulfilment and individual emancipation - to materialistic - security of material value.
The theory consists in saying that this change is due to the fact that generations after the Second World War were socialized in a situation referring to two factors :
- economic growth
- the expansion of the welfare state
According to this theory, people socialized during the time of expansion developed post-materialistic needs because they did not have the need for security; hence there is a tendency to value scarce resources.
There is another element that is focused on the idea of scarcity of certain resources, as people tend to favour resources that are scarce (economic wealth was daily), as they were socialized during the period of wealth expansion. On the other hand, this difference is greater in countries that have had greater economic expansion.
In this case, one cannot yet test or falsify the theory, one must move from theory to hypotheses by going towards something a little more specific which makes it possible to corroborate the affirmations :
- young people are more post-materialistic than older people in Western countries: we are interested in young people compared to older people, the hypothesis is tested if young people are more post-materialistic than older people.
- the difference between young people is less young is greater in countries where the change in quality of life has been stronger, in other words in countries where economic expansion at that time was the most important as in Germany.
- Post-materialists are more numerous in the richest countries; indicators can easily be found to test this hypothesis.
There has been a set of organically articulated proposals, however this is useless to test the theory; for this it is necessary to formulate hypotheses that are also at the abstract and theoretical level.
It is a theory of political behaviour also known as the "Michigan model" proposed in the 1950s. This theory postulates that people vote because they feel loyal to certain parties; it is through partisan identification that people will vote for a party because they identify with it.
This sense of party identification comes from the socialization process. However, we do not have enough substance, we are at the level of a set of proposals linked together in an organic way.
First of all, it is necessary to specify the hypotheses which are for example:
- people who identify with the Socialist Party tend to vote for the Socialist Party; the degree of abstraction has fallen a notch.
- People from working-class backgrounds tend to vote for the socialist party: this is a testable hypothesis, because one can easily go to the field to collect data.
Theory of political opportunities[modifier | modifier le wikicode]
This theory says that people mobilize because they are unhappy or because there are certain political opportunities to take to the streets.
The theory says that the forms and levels of mobilization depend on political opportunity structures. These political opportunities are to be sought in the structure of the state and in the degree of openness and closure of the state:
- The demonstrations are smaller and at the same time they are more radical and violent in countries characterized by closed opportunity structures. We can test this hypothesis because we can identify more open or closed states.
- the more the police repress demonstrations, the more radical they tend to become. It is enough to observe manifestations: in this case, there is a problem of endogeneity which is the problem of reverse causality, because the hypothesis postulates that the more the police will repress the more there will be a tendency to radicalisation, however the relationship could be the opposite and we do not know what explains what.
Operationalization[modifier | modifier le wikicode]
Definition of operationalization[modifier | modifier le wikicode]
We will focus on the moment when we pass to the field; operationalization is the moment when we define the research drawing: we start from a given theoretical framework and then we go to the field, we will deal with this passage.
In order to be able to control and verify, by taking up the idea of the critical theory i.e. to be able to falsify a hypothesis, one must be able to set up certain passages which answer the name of operationalization. This is a key moment in the research process.
Let us recall that according to Corbetta, the concept refers to the semantic content, therefore to the meaning of linguistic signs and mental images; the concept is an abstraction of reality, it is basically something general. In other words, the only way to know and think about a reality is conceptualization, which is the foundation, a fundamental phase of each scientific discipline.
On the other hand a concept can refer to abstract and non-observable mental constructions such as the concept of power or the social class, a concept can also refer to more concrete and observable entities such as a chair or a worker however a concept always refers to the class of objects.
It is through the realization of concepts that one can establish an empiry.
Phases of operationalization[modifier | modifier le wikicode]
It can be divided into several phases, these are the key moments in the research process:
1) Render concepts into object properties (unit of analysis): concepts must be assigned and applied to objects; they are units of analysis that refer to the choice of analysis on which we will work. In other words, it means moving from a conceptual level to a measurable empirical level, transforming concepts, applying them to concrete objects and thus to units of analysis.
- the concept of "power": must be able to be transformed into an object, for example, the role of power in an enterprise: we begin by defining the unit of analysis.
- Economic development: must be applied to something concrete that could be the concept applied to nations.
- electoral participation: one may be interested in territorial units that are collective property or individual property such as the frequency of participation in demonstrations.
Nota bene : these properties of objects will have different states depending on the objects in question, for example the economic development of France differs from that of Switzerland. The object properties vary according to the selected criterion.
2) To give an operational definition of concepts: rules must be established and decided in order to translate these concepts into empirical operations, in other words, rules must be established to translate the concepts into empirical operations.
- power concept: power is defined first as the role you can have in an organization. Then we must specify the number of people over whom the individual exercises power (he can direct 1000 or 100).
- voter turnout: if it is assumed that voter turnout is measured at the commune or canton level, then the percentage of voters in relation to the number of voters should be considered.
3) Apply the operational definition to concrete cases: this is the operationalization phase in the strict sense of the term, we are going into the field which makes it possible to define a variable.
- variable: it is the result of the process, we move from a concept to a variable, they are the theoretical concretization of a concept.
- modality: a value is applied to each modality as, for example, for the concept of level of education: university 5, primary 1, etc. This makes it possible to assess a person's level of education by establishing a code.
Therefore, operationalization in the strict sense of the term is the transition from the property (concept) to the variable that depends on how the transition is made so that one can have different variables.
Operationalization depends on how the concepts are translated:
- counting (counting units)
It is necessary to reflect on what type of analysis we want to lead through the development of concepts.
It is essential to define the concepts; the concept has a relationship of meaning, it is a fundamental element of scientific research.
- Weight: weight of a book (1 kilo): it has no relation between the physical weight of a book and its impact in literature.
- Age: Age of a person (20 years).
- Education: level of education (university).
- Power: political role (MP, minister, senator): it is difficult to define who has the most power, these are roles in which we cannot establish a hierarchy.
- Participation: Voting (frequency).
Unités d’analyse[modifier | modifier le wikicode]
Dans la recherche empirique, on doit définir des unités d’analyse. L’unité d’analyse représente l'objet social ou politique dans la recherche empirique, c'est essentiel de la définir.
On distingue trois niveaux d’analyse, mais qui dépendent du contexte de la recherche :
- macro ;
- méso ;
On peut approfondir la distinction à 6 niveaux d’analyse :
- individu : ce sont les personnes.
- agrégat d’individus : ce sont des variables collectives agrégées ; c’est l'ensemble des individus qui est une variable collective. Par exemple si le taux de participation en Suisse est de 40%, ce calcul est effectué sur la base des variables individuelles.
- groupe / organisation / institution : variables collectives et structurelles, on ne passe pas par une agrégation des comportements individuels, c’est un processus différence de l'agrégat.
- événement : par exemple dans les études faites sur les révolutions, chacune peut être divisée en sous-événements.
- produit culturel : par exemple un tableau qui permet d’expliquer l’évolution d’une branche artistique.
- relation : accords, collaborations, des relations organisationnelles ou interindividuelles.
The first three levels of analysis are the most frequent, aggregated variables. There are structural or global variables that characterize an individual or group as such. Aggregate variables are derived from mathematical operations on individual variables whose unit of observation is at a lower level while structural variables are at the unit of analysis level.
At the end of the process, we have "cases" that are copies of a given analysis included in a research when we speak of unit of analysis it is an abstract or general case, while the "case" is concrete is multiple to know what we are going to study.
Thus "cases" are the specific objects of research that can be defined once one has moved from the steps of defining a concept to the variables that allow one to choose cases and see how they vary on the variable resulting from the process.
There is really no right or wrong working definition, it is a matter of being as explicit and transparent as possible. Therefore, the choice made during the operationalization phase must be explained and justified.
There is always a gap between the empirical and theoretical level, one can never arrive at the perfect identification that can arrive at a fair or false operative definition.
Finally the danger in this phase is not in the reduction that is inevitable, but in the commodification, that is, in identifying the concept with the variable.
The operational definition must meet criteria of objectivity, it must arrive at a controllable process that can be repeated by others.
To know if it is a good operationalization, a choice must be justified, that is, it must meet a criterion of objectivity and justification without eliminating arbitrariness.
Criteria for distinguishing variables[modifier | modifier le wikicode]
The variable is an operationalized concept; there are several ways to define variables and therefore several ways to classify them:
- non-manipulable / manipulable:
- non-manipulable: these are variables that cannot be modified, for example, socio-demographic characteristics.
- manipulable: the questions to ask.
- dependent / independent:
- dependent: variables explained; this is what we want to explain also called endogenous variable.
- independent: explanatory variables; it is supposed to explain also called exogenous variable.
- unobserved (latent) / observed (manifest):
- unobserved: values are unobservable latent variables.
- observed: opinions can be observed, for example.
Nota bene : when we work on values in science-politics we approach attitudes through we go back to something unobservable.
- Individual / collective (aggregate, global, contextual)
- Value processing: this is the most important aspect, it is related to measurement. There are different types of variables. Knowing what type of variable we have to do will tell us what type of analysis we have to do; the whole process of operationalization and the end of the process, namely the creation of variables is fundamental leading to variables of different natures.
Types of variables[modifier | modifier le wikicode]
There are three types of variables that can be distinguished between four criteria
|Property states: the values of the variable||
Non-ordered and non-ordinable categories.
ex : nationality, religion
Also categorical, but ordered; one can create an order.
e.g. education level, to what extent are we interested in something.
More categories, but variables :
We can put them in a certain order. They can be included in different categories. There is an ordonnénement, the ordinal variable results from the operative definition which consists in giving an order to the various objects.
interval between them is the same (1 year, for example) -v. continuous-
we can count them -v. discreet-
The characteristic of values is names.
ex : Canadian, Swiss
The categories must be exhaustive. All categories must be contemplated and mutually exclusive.
Number with ordinal properties.
ex: little, enough, very; we associate a number to each state, this code is arbitrary.
Number with cardinal properties, the number reflects a real property.
ex: went to vote 5 times, one cannot arbitrarily associate figures.
|Operations that can be carried out on securities||
Equality or inequality.
ex: Muslim differs from Catholic
Equality or inequality, higher or lower order.
All mathematical operations, equivalences, differences, multiplication, etc. can be applied.
ex: a 40-year-old is twice a 20-year-old.
Corbetta distinguishes the quasi-cardinal variables, they are situated between the two, namely between the ordinal and the cardinal. They would be ordinal variables, but they are considered to be cardinal variables. We try to make a discrete or ordinal variable continuous.
These are ordinal variables that we try to render as continuous variables. We try to compare the difference between two values, for example (not, little, enough, very); we cannot say that the distance between "not" and "little" is the same as between "enough" and "very". They can be ordered, but the distance cannot be measured.
One way to proceed are the scales, for example from 0 - 10 to define whether we are left or right. From then on, we pass from ordinal to cardinal variables.
Relationship between concepts and indicators[modifier | modifier le wikicode]
It is the operationalization of complex concepts. Generally, complex concepts are not observable, one can only observe their manifestation, for example deviance, religion, power. These concepts are at a higher level of generality and abstract, and cannot be observed directly.
Most concepts in the social sciences can be defined as complex concepts that are more difficult to operationalize, that is, to transform them into the property of a unit of analysis.
Example: concept of religiosity; five different definitions are used to formulate it which are increasingly specific:
- believing in a divinity: allows us to move towards concretization.
- believing in the Christian god: each religion has its own definition of god.
- belong to the Catholic Church
- act according to the rules of the church: higher degrees of precision.
- go to church every Sunday: one can try to operationalize the concept of religiosity by reducing it to going to church every Sunday.
Thus, we see how we can pass from the general to the specific through different passages.
How to measure and operationalize these complex concepts?
The concept can be subdivided into sub-concepts called indicators. Indicators are crucial in the operationalization process.
An indicator is a simpler, more specific concept of the original concept that can be immediately translated into observable terms.
Indicators are linked to more general concepts by an indicative relationship between the concept and the indicator. We go down the generality scale to more specific concepts; it is a semantic representation relationship between the indicator and the concept it is supposed to represent, indicate, measure.
In other words, we go down the scale of generality and abstraction from general concepts to more specific concepts linked to the first by affinities of meanings.
Nota bene : there is no right choice of indicators.
What is the relationship between the concepts and the indicator?[modifier | modifier le wikicode]
A concept cannot be captured entirely by a single indicator, a given indicator covers only one aspect of that complexity of the concept. The indicators are partial representations. It is always necessary, if possible, to find several indicators for the same complex concept a same complex concept can never be indicated by only one indicator, there is a criterion of multiplicity of indicators.
Example - religious practice may be an indicator of the component of the ritual dimension of religiosity, but religiosity also has other components such as religious feelings, religious ideology, religious affiliation, etc.
One should always be aware that an indicator is always in a biased relationship with the general concept it is supposed to indicate.
An indicator may overlap only partially with a concept in other words the same indicator may be linked to several concepts, it may indicate, mean, represent different concepts.
Example - in theocratic societies, religious practice can be an indicator of social conformity rather than religiosity. The practice of religion can be both an indicator of social conformism and religiosity.
The same indicator only partially covers a concept while being an indicator of different concepts.
The choice of indicators is arbitrary, so they should be argued rather than shown to be correct. We must try to show the close link between the theoretical dimension of the concept and the empirical dimension, the two things cannot be dissociated.
The indicators of a complex concept can be found in several ways according to logical reasoning and even according to common sense or more systematic according to what has been done in previous research with an importance of literature.
Empirical translation of complex concepts: phases of the operationalization of complex concepts[modifier | modifier le wikicode]
If we have concepts that are not multidimensional, this phase can be suppressed; if we work with a complex concept, we must first simplify the complex by going through dimensions, it is a theoretical reflection, we analyze the concept in its main components of meanings.
There are four phases:
- Articulation of the concept in dimensions: we reflect on the other dimensions of the concept such as, for example, religiosity which has dimensions of practices, ideologies, etc. It is the passage from general abstractions to specific, we say that we can divide it into sub-concepts for each dimension, however we are not yet in the operationalization phase. The different aspects and meanings of the concepts are questioned.
- Choice of indicators: we ask ourselves the question of empirical translation, we decide which indicators we will choose. Indicators are more specific concepts, we are beginning to take a step towards variables such as being interested in participation in rituals which is part of the ritual participation or the religious practical dimension.
- Operationalization: indicators that are still concepts are transformed into variables. It is the creation of variables that can be, ordinal, cardinal or intervals. For example, with regard to religious practice, we will measure and operationalize religious practice, which is an indicator of one dimension of religiosity, that is, the number of times we go to church each year. This indicator falls under the behavioural component because choosing a "practical" indicator of religion determines a frequency.
- Index formation: all indicators are synthesized into a global measure. We are going to proceed with the formation of the indices, we are trying to group these indicators under a single measure, perhaps for example the construction of scales; on the empirical, concrete, specific level, we are trying to arrive at a measure, because it is easier to work with a variable than with a multitude of variables.
Depending on the research objectives, several measures or indicators will be selected.
This graph shows the process that goes from the complex concept to indicators or more specific indicators that indicate the concept; then variables were created and then in the last step we will group the variables into a single measure called the index.
Through this operationalization process variables are created that can be, ordinal, cardinal, categorical or interval variables - ordinal. In this example we would have nine indicators from which to construct an index that summarizes the concept. One starts from a concept which is the theoretical level towards a variable, the index is a variable derived from the sum of the other operations on the various variables.
In this process there is always a possibility that there are errors that get introduced so that a variable is never completely assimilable to concepts, there is always a lag what is important is first to know what are the different sources of errors that produce the lag.
Some errors can be corrected and others cannot, but knowing the problem is something very important.
Empirical translation of complex concepts: examples[modifier | modifier le wikicode]
Nota bene : we started to specify the concept through seven dimensions
Nota bene : the distinction between concept and dimension is relative, now participation has become a dimension of another concept notably through the criteria of polysemy, partiality and arbitrariness.
The complex concepts and indicators are all at the bottom of the concepts, we enter the empirical phase with the last step.
The idea is that we move from an abstract and general concept through sub-dimensions that allow us to choose good indicators in the sense that they are justifiable and justified in the context of theory being an indicative relationship with the concept we want to measure.
Errors in the transition from concepts to indicators[modifier | modifier le wikicode]
There is only partial coverage of the concept by the indicator, but there is always a gap between the observed value and the true value related to the concept being measured.
First, a distinction must be made between two types of errors:
- systematic error, "constant error
- accidental error, "variable error
Total error is the sum of accidental error and systematic error.
TOTAL ERROR = ACCIDENTAL ERROR + SYSTEMATIC ERROR
- systematic error
Occur in all our measurements in a systematic way, for example everyone tends to overestimate their own participation.
- accidental error
It is a variable error from one measurement to another, we measure differently at different times.
One of the two types of errors is more easily detectable than the other. If a problem remains constant, if we do not assume problems then we will not notice anything, that is why an accidental error is more easily identifiable.
There can be different types of errors resulting from a distinction between two phases:
1) indication phase: theoretical
Two types of errors can be distinguished:
- Indicator: Due to poor indicator selection, the indicator does not measure what it is intended to measure. It is an error that is almost by definition constant or systematic, difficult to detect except through logical reasoning and intuition. In this case there is a problem with the validity of the indicator, i.e. it does not really measure the concept it is supposed to measure.
- systematic: once done wrong, it will have repercussions on research.
2) empirical phase: operationalization error These errors can result from operationalization errors, they can be systematic or accidental. We can distinguish three sources of operationalization errors, in other words there are three moments at which we are subject to danger and risk of falling into these errors. However we will ignore the data processing error.
- case selection: badly chosen cases, there may be errors.
- coverage: consists of the fact that we did not cover the population we wanted to cover.
- sampling: if the sample is taken according to certain procedures, the percentage of error can be calculated.
- non-response: there are individuals who do not wish to respond to the survey, which will bias the analysis.
There are several sources of errors related to the selection of subjects in a first phase which is also an operationalization error.
- observation: poor observation of cases
- interviewer: errors linked to the interviewer, he/she could subject the interviewee to direct or indirect pressure.
- interviewee: the person may misunderstand the question or deliberately bias the research.
- instrument: the manner in which the issue is administered.
- data processing mode: analyse in the wrong way.
There can be errors that make the initial concept no longer correspond or only partly correspond with the final concept which is the variable. To do this, we must be aware and put in place everything possible to reduce the time lag as much as possible. Note that the only type of error that can be measured is sampling error.
When we analyze the data, when we have the variables, we must be aware that the variable is only an approximation of the concept it is supposed to measure or operationalize.
On the other hand, we must ensure that these sources of errors are reduced to a minimum by trying to avoid any bias linked to the person interviewed, by using the right instrument and the right method of administration while covering the entire population that we are supposed to study by reducing non-response.
Reliability and validity[modifier | modifier le wikicode]
Indicators may be more or less reliable and valid. The question is to what extent is a "measure" reliable and valid?
The notion of reliability refers to the possibility of reproducing the same measurement, i.e. the reproducibility of the measurement. It is the degree to which a certain procedure of translating a concept into a variable produces the same results in repeated tests with the same measuring instrument (stability) or with equivalent instruments (equivalence).
On the other hand, there is reliability related to internal consistency when there is a series of variables that are supposed to be part of the same concept or to measure the same concept. In this case there are coefficients which make it possible to measure this reliability as the alpha of Cronbar.
It is an Adequacy, the degree to which a certain procedure of translating a concept into a variable actually measures the concept that one intends to measure. A valid indicator is one that really measures what you want to measure.
To the question to what extent the variable we have operationalized, captured, captures and measures the concept as well as the reality we want to discover, we must refer to an adequacy.
Research aims to find indicators that are both reliable and valid.
References[modifier | modifier le wikicode]
- Page personnelle de Marco Giugni sur le site de l'Université de Genève
- R. Boudon; Les méthodes en sociologie, p.31