Analyse de données qualitatives

De EduTech Wiki
Aller à : navigation, rechercher
Manuel de recherche en technologie éducative
Page d'entrée du module
◀▬ Analyse de données qualitatives ▬▶
brouillon intermédiaire
2019/01/23 ⚒⚒ 2015/08/27
Contenu du module

  • Qualité: brouillon
  • Difficulté: intermédiaire

1 Introduction

Ce module présente différents aspects de l’analyse des données qualitatives. Nous allons présenter une approche structurelle «moderne» qui requiert du chercheur de coder les données. Ces codes lui permettront alors de mener divers types d’analyses, dont nous allons vous montrer quelques exemples.

Objectifs d’apprentissage

  • Apprendre à coder des données et créer des manuels de codage (codebooks)
  • Apprendre les fondements de quelques techniques d’analyse descriptive (notamment les situations et les rôles)
  • Apprendre les fondements de quelques techniques d’analyse causale

Caractéristiques d'un chercheur qualitatif

"un bon chercheur qualitatif est doté des caractéristiques suivantes:

  • Une certaine familiarité avec le phénomène et le milieu étudiés,
  • Un intérêt affirmé pour la dimension conceptuelle,
  • Une approche pluridisciplinaire par opposition à une formation restreinte ou cantonnée à une seule discipline,
  • De solides qualités « d’investigateur », comprenant de l’obstination, la capacité à faire parler les gens, et la capacité à prévenir une clôture prématurée".
Miles, M. & Huberman, M. (2003, p. 78). Analyse des données qualitatives. 2e édition. De Boeck Université.

2 Les différentes méthodes d'analyse de données qualitatives

2.1 Les méthodes non spécialisées

L'extrait suivant est tiré de Savin-Badin, M. & Howell Major, C. (2013, pp. 434-440). Qualitative research. The essential guide to theory and practice. London: Routledge. "Data analysis is a systematic search for meaning. It is a way to process qualitative data so that what has been learned can be communicated to others. Analysis means organising and interrogating data in ways that allow researchers to see patterns, identify themes, discover relationships, develop explanations, make interpretations, mount critiques, or generate theories. It often involves synthesis, evaluation, interpretation, categorisation, hypothesising, comparison, and pattern finding. It always involves what Wolcott calls « mindwork »…. Researchers always engage their own intellectual capacities to make sense of qualitative data”. (Hatch, 2002, p. 148) (…)

There are a number of analytical approaches from which qualitative researchers may choose. Which method to choose is a critical decision, because the method invariably influences the focus of data analysis and thus the results. (…)

A keyword analysis is exactly what it sounds like. It involves searching out words that have some sort of meaning in the larger context of the data. The idea is that, in order to understand what the participants say, it is important to look at the words with which they communicate. (…)

Constant comparison is an analytical method that researchers use to develop themes and ultimately generate theory. (…) The constant comparison process involves the following basic steps :

  • Identifiy categories in events and behavior.
  • Name indicators in passage and code them (open coding)
  • Continually compare codes and passages to those already coded to find consistencies and differences
  • Examine consistencies or patterns between codes to reveal categories
  • Continue the process until the category « saturates » and no new codes related to it are identified.
  • Determine which categories are the central focus (axial categories) and, thus, form the core category. (…)

Content analysis is the process of examining a text at its most fundamental level : the content. It is an analysis of the frequency and patterns of use of terms or phrases and has been applied to a range of research approaches. While content analysis typically has been applied to print/text, it also has been adapter to analyze visual artefacts. Over time, this mtehod of qualitative data analysis has taken two primary forms: classical content analysis and ethnographic content analysis.

Classical content analysis: Classical content analysis strives to make "replicable and valid inferences from data to their context (Krippendorf 1080:21) as well as to make "objective, systematic, and quantitative descriptions of the manifest content of communication" (Berelson 1952:489). It also seeks to "make inferences by systematically and objectively identifying specified characteristics within text" (Stone et al 1966:5). Classical content analysis involves the following steps:

  • examine the text or aretfact
  • peruse it in its entirety
  • determeine the properties of the text/artefact
  • examine overt and latent emphases
  • determine rules for categorizing:
  1. determine how much data will be analyzed at any one time (whether a line of text, a sentence, a phrase, part of the artefact)
  2. determine categories
  3. they must be inclusive (all examples fit within the category)
  4. they must be mutually exclusive (no examples fit within more than one category)
  5. determine thees that emerge from the categories.

Domain analysis is a collection of categories that are related. For example the domain of chocolate would include sweet, bitter and semisweet. (…)

Thematic analysis is a method of identyfing, analyzing and reporting patterns in the data (Braun and Clarke 2006). There is no clear agreement for what thematic analysis is or how one does it, although it appears that much of what qualitative researchers do when analyzing data under the generalist term of qualitative data analysis is actually thematic analysis (Braun and Wilkinson 2003). (…) The method provides a general sense of the information through repeated handling of the data. The idea is to get a feel for the whole text by living with it prior to any cutting or coding. It is not the most scientific sounding method but we believe it to be one of the best. The researcher can rely on intuition and sensing, rather than being bound by hard and fast rules of analysis. Braun and Clarke (2006) recommend doing the following when conducting thematic analysis:

  • familiarize yourself with your data
  • generate initial codes
  • search for themes
  • review themes
  • define and name themes
  • produce the report.

What is unique about thematic analysis is that it acknowledges that analysis happens at an intuitive level. It is through the process of immersion in data and considering connections and interconnections between codes, concepts and themes that an « aha » moment happens."

2.2 Les méthodes spécialisées

L'extrait suivant est tiré de Savin-Badin, M. & Howell Major, C. (2013, pp. 440-447). Qualitative research. The essential guide to theory and practice. London: Routledge. "Qualitative researchers also have an array of more specialised methods of data analysis that tend to be paired with a specific philosophical position and one or two research approaches. These analytical methods tend to require a high level of skill on the part of the analyst, extensive reading to develop knowledge of the method and, at times, expert guidance. (…)

Analytical induction : The process of analytical induction requires an examination of similarities between phenomena or events for the purpose of developing basic concepts of understanding. Analytical induction proceeds with an investigation of broad categories of understanding and then it moves on to developing sub-categories. It has been one of the classic research methods associated with ethnography. The basic process is as follows :

  • Examine the event.
  • Develop a hypothetical statement of what happened during the event.
  • Examine a different but similar event to determine whether the new event fits the hypothesis.
  • If the event does not fit the hypothesis, revise the hypothesis, repeat the process and revise the hypothesis until it explains all examples encountered.
  • Develop a hypothesis that accounts for all cases.

This process often is used in ethnography. (…)

Heuristic or phenomenological analysis: The word heuristic comes from the Greek heuriskein, meaing “to find, to discover”. This approach goes hand in hand with phenomenography (…) and requires an attempt to discover how an individual in a context makes sense of a particular phenomenon. It needs a combination of psychology and interpretation and generally may be viewed as a problem-solving approach to data analysis. Phenomena are considered generally to relate to an occasion of individual significance, such as a major life event. (…)

Hermeneutical analysis : The term hermeneutic comes from the Greek hermeneuein, which means « to interpret ». This method (…) in the social science context involves the analysis and interpretation of social events and their meanings to participants. This method also has been most often paired with phenomenology / phenomenography. The emphasis is on context and behaviour, the purpose being to interpret the general meaning in the context in which it occurs. It is a consideration of the relationship of the whole to the parts and the parts to the whole. Meaning is thought to reside in author intent/purpose, the context and the encounter between author and reader. Themes are related to the dialectical context. The process involves:

  • seeking the meaning of text for people in the situation
  • telling the participant’s story
  • bracketing the self out in analysis (to avoid telling the researcher story rather than the researched story)
  • seeking to interpret different layers of text
  • constructing knowledge by using context to understand and create. (…)

Ethnographic analysis, true to its name, goes hand in hand with ethnography. (…) There is no one method of ethnography but rather, there is a set of general strategies that ethnographic researchers tend to use. According to Merriam (2009), the ethnographer uses a classification scheme, either based upon concepts typically found in culture (emic perspective) or developed by the researcher (etic perspective). (…)

Narrative analysis is not one specific method but rather is a range of methods that frequently are used with the narrative approaches that we describe in Chapter 15 . (…) In qualitative research, narrative analysis requires the researcher to focus on the ways in which participants use stories to interpret the world. It treats stories as interpretive, « storied », social products that individuals produce in unique contexts, to represent themselves or their worlds, rather than as facts to be assessed for « truthfulness ». Interviews are thus viewed as « storied » and necessarily biased. Three key ways to analyze narrative influence are :

  • structural analysis, which focuses on core events (see Labov 1973)
  • sociology of stories, which focuses on cultural, historical and political contexts (for example, Plummer 2001)
  • functional analysis (Bruner 1990), which focuses on what work stories do in participants’ lives. (…)

Discourse analysis is not a specific method but rather is a term that describes a range of methods of analyzing language, whether through text, speech or sign (Burman & Parker 1993 ; Potter & Wetherell 1987 ; Willing 2003). It has been at times paired with ethnography, narrative approaches and case studies, as well as action research approaches. It involves the « linguistic analysis of naturally occurring connected spoken or written discourse » (Stubbs 1983) and provides « insight into the forms and mechanisms of human communication and verbal interaction » (Van Dijk 1985). The purpose is for the construction and negociation of power and meaning in discourse and interaction. The object of analysis is any communication event. The event is described as a set of speech acts, such as sentences, rhetorical devices, turn taking, conflicts, truth claims or propositions. The preference is to analyze language as it occurs in natural text, although focus groups interviews are often analyzed too. Moreover, discourse analysts have analyzed doctor-patient interaction, police interactions, court proceedings and a host of other interactions. The general process involves :

  • a search for patterns (questions and answers, who dominates the discourse and how, or any other observable patterns of interaction)
  • an attempt to « map out » discourse structure and function as well as the relationships between participating individuals
  • engagement in close analysis of language.

Some of the most influential developers of these approaches include Gee (2005) and Halliday and Matthiessen (2004). (…)

Semiotic analysis involves a study of signs and symbols as well as how their meaning is constructed within a culture. It has been paired with arts-based approaches, narrative approaches and ethnography. This kind of analysis involves a broad view of cultural products including popular media, such as digital and visual artefacts. This method assumes that meaning is not inherent in the products ; rather, their meaning is derived through their relationships with other things. There are three critical factors that warrant consideration: the sign vehicle, sense and referent. Semiotic analysis involves the following :

  • identification of the text or object
  • examination of the researcher’s purpose in selecting the text or object
  • clarification of the sign vehicle
  • description of modality (reality claims)
  • analysis of paradigm (for example, genre, theme)
  • consideration of what is termed the syntagmatic structure of the text (such as narrative argument)
  • examination of rhetorical tropes (such as metaphors or metonyms)
  • examination of intertextuality (for example, does it allude to other texts or genres)
  • analysis of semiotic codes (representations).

Event analysis : The purpose of an event analysis is to examine and represent the chronological series of events in an actual or folkloristic accounting as logical structures (Tesch 1990). The idea is that people cause or prevent events from occurring, which can provide important evidence. It begins with identification of a specific starting point and a specific end point. An abstract logical structure of the event is developed and then compared with the actual event. The analysis examines elements and the connection of elements, including boundaries, as well as the assumptions that govern the connections. The goal is to develop an explanatory model (Heise 1988). This method is often used with video, where events can be watched again and again. A related method is critical event analysis, which involves presenting participants with typical scenarios of behaviour. The researcher sollicits the following from participants :

  • opinions about the cause
  • structure and outcome of an incident
  • information about participant feelings and perceptions
  • description by participants of actions that were (or should have been) taken during the incident
  • resultant changes they might see in their own future behaviour. (…)"

2.3 Comment choisir une méthode d'analyse de données?

Avec toutes ces méthodes d’analyse de données, il est difficile de savoir laquelle choisir. Pour vous guider, votre question de recherche générale ainsi que l'objectif de votre recherche vous permettront de savoir précisément ce que vous recherchez et d'évaluer si vous allez parvenir au résultat escompté avec une méthode d'analyse donnée. Pour choisir, étudiez soigneusement les différentes possibilités et expliquez, de manière argumentée, comment vous arrêtez votre choix.

3 Principes de l'analyse de données qualitatives

Avec l’analyse qualitative, le chercheur essaye d'identifier une structure dans les données (comme le font les techniques exploratoires quantitatives). Pour ce faire, deux types de techniques d’analyse sont couramment utilisés:

  1. Une matrice est une tableau qui engage au moins une variable, e.g.
    • Les tableaux de variables centrales selon les cas (équivalents aux statistiques descriptives simples telles que les histogrammes)
    • Les tableaux croisés permettant d’analyser comment deux variables interagissent
  2. Un graphique (réseau ou carte conceptuelle) permet de visualiser les liens entre les données:
    • liens temporels entre des événements
    • liens de causalité entre plusieurs variables
    • diagrammes d'activités et de processus
    • etc.
Analyse = mise en tableaux et visualisations diverses des données

L’analyse des données qualitatives comprend généralement une série d’étapes itératives liées. Le principe général de la plupart des méthodes d’analyse des données qualitatives est le suivant:

  1. Les données doivent être codées et indexées pour pouvoir être retrouvées pour l’analyse. Plus précisément, le codage d’informations permet d’identifier les variables et les valeurs. Une telle analyse systématique des données augmente la fiabilité et la validité de construction, i.e. vous devez observer tout ce qui permet de mesurer les concepts.
  2. Vous devez ensuite créer des visualisations, des matrices, des grammaires, etc. pour interpréter les données.
  3. Vous devez ensuite interpréter, faire émerger du sens, de ces visualisations.
  4. Vous devez finalement vérifier la pertinences de vos analyses et interprétations.
L’analyse des données qualitatives - résumée

Quelques conseils:

  • Lorsque vous utilisez ces techniques, gardez toujours un lien avec la source (autrement dit, les données codées).
  • Efforcez-vous de faire rentrer chaque matrice ou graphique dans une seule page (ou assurez-vous de pouvoir imprimer les travaux réalisés à l’aide d’un ordinateur sur une page A3) afin d'avoir une vue d'ensemble de toutes les données.
  • Privilégiez une vision synthétique, mais préservez suffisamment de détails pour rendre vos artefacts interprétables.
  • Consultez des manuels spécialisés e.g. Miles, Huberman & Saldaña (2014) pour des procédures validées et/ou inspirez-vous de travaux de recherche qualitative publiés dans votre domaine.

4 Avant de commencer le codage

Avant d'expliquer le codage et l’analyse, prévoyez un système de gestion de vos données. Pour plus d'information sur la gestion des données actives de recherche, voir notamment le point 4 de la page dédiée.

Par ailleurs, avant et pendant votre analyse, pensez à:

  1. Rédigez des mémos pour conserver une trace de vos intuitions. Il est utile d’écrire des petits mémos lorsqu’une idée intéressante surgit à la suite d’une observation lors de l'analyse des données.
  2. Créez des fiches de contacts et/ou des fiches de synthèse qui vous permettront de garder en tête votre travail de terrain. Après chaque contact (téléphone, interview, observation, etc.), rédigez un document bref qui devrait inclure:
    • Une étiquette claire pour des raisons d’indexation (nom de fichier), e.g. Contact_Participant1_2005_3_25.doc.
    • Type de contact, date, lieu, et lien vers les notes de l’entretien, transcriptions.
    • Thèmes principaux abordés et variables de recherche traitées (ou renvoi vers la page de l’entretien).
    • Premières remarques interprétatives, spéculations nouvelles, éléments à traiter dans un deuxième temps.
    • Questions à traiter lors de l'entretien suivant.

5 Différence entre résultats et recherche en cours

Avant de commencer votre analyse, réfléchissez bien à ce dont vous avez besoin (échantillon) pour pouvoir répondre à vos questions de recherche et pensez à consulter des ouvrages de référence pour choisir la méthode la plus appropriée. Remarque: dans le cas de nombreuses études qualitatives rapportées sous forme d'articles dans la littérature, vous remarquerez que les chercheurs présentent souvent uniquement des citations d’entretiens. Ces citations sont choisies pour représenter des opinions spécifiques et sont arrangées selon un ordre logique, e.g. des sujets émergeant dans la perception de l’utilisateur sur des problématiques données. Cependant, avant de rédiger leur article, ces chercheurs ont utilisé des techniques d’analyse comme celles mentionnées dans l'ouvrage de Savin-Baden & Howell Major (2013, pp. 434-447).