An introduction to educational research methods. Введение в образовательные исследовательские методы Білім беру-зерттеу әдістеріне кіріспе

Analysing Qualitative Data

421

Матрицалар сонымен бірге пәндік зерттеу форматы үшін сәйкес келеді, сондықтан 

кодталған мәліметтер жеке тұлғалар үшін немесе басқа талдау бөлшектері үшін, 8.2 

кестеде келтірілгенге ұқсас топтарға емес, берілуі мүмкін. Бұл кесте жыныс сияқты 

тәуелсіз айнымалылар негізінде түзілген. 8.2 кестеде осы тарауда бұрын келтірілген 

оқушылардың онлайндық өзара әрекетін зерттеу үшін NVivo бағдарламасының 

көмегімен түзген матрицаның бөлігі келтірілген. Мәліметтерді ұйымдастыру тізімі 

келтірілген тақырыптарға кодталған жауаптарда маған жыныс айырмашылығын 

табуға мүмкіндік берді: атап айтқанда, матрица хабарламалар тақтасын пайдалану 

түрлі оқыту түрлеріне әкелетіні туралы оқушылардың түсініктемелеріне жасаған 

менің талдауыммен бекітіледі. NVivo сияқты бағдарламамен түзілген матрицаның 

айырмашылығы – таңдап алынған сілтемелерге деген мүмкіндікке бірнеше ұяшыққа 

басу арқылы алуға болады. Қолмен жасалған матрицаларда ұяшықтар сұхбатта 

келтірілген негізгі тақырыптарды білдіретін мәлімдемелерді (осында көрсетілгендей 

сандардың орнына) қамтуы мүмкін. Осылайша, зерттеуші негізгі туындайтын 

дәлелдемелерге аралық ауызша шолу жасау арқылы ұяшықтарды толтыруы тиіс. 

8.1 тапсырма

Осы тараудағы жазылымнан алынған үзіндіні тағы бір рет оқып шығып, оның 

көмегімен келесі ашық кодтарды кодтаңыз:

 

 ағылшын тілінде сөйлесу

 

 француз тілінде сөйлесу

 

 оң қарым-қатынас

 

 теріс қарым-қатынас

 

 тілдік қиындықтар

Қандай  басқа  концепциялар  үзіндіден  туындайды  және  сіз  оларды  қалай 

кодтар едіңіз? 

Қандай in-vivo кодын зерттеуші осы үзіндіден таба алады? 

Негізгі ойлар

Бұл  тарауда  сапалық  мәліметтерді  талдау  үдерісінің  кейбір  негізгі 

ерекшеліктерін келтіріп отырмын. Мен сонымен бірге зерттеуші педагог үшін 

ерекше пайдалы болуы мүмкін тәжірибелік құралдар мен стратегияларды 

Analysing Qualitative Data

422

Рефлексияға арналған сұрақтар

1  Мәліметтерді  тіркеу  үшін  қандай  құралдарды  пайдаланасыз  және 

мәліметтерді талдауға қалай қол жетімді етесіз? 

2  Шетке  шығып,  белгілі  нәрсені  белгісіз  етіңіз.  Жазылым  сізбен  сөйлесуі 

керек; қойылған сұрақтар мен мәселелерге аса назар аудармаңыз. 

3  Сапалық  талдау  бағдарламасының  сіз  үшін  талдау  жүргізбейтінін  есте 

сақтаңыз.  Сіз  мұны  орындауыңыз  қажет.  Компьютерлік  бағдарлама  өз 

мәліметтеріңізді  оңайырақ  ұйымдастыруға  мүмкіндік  береді,  өйткені 

мәліметтердің белгілі бір бөліктеріне топталып, шоғырланған болуы тиіс. 

4  Сапалық  зерттеудің  шынайылыққа  жеткізбейтінін  есте  сақтаңыз.  Өз 

ойларыңызды  абайлап  келтіргеніңіз  дұрыс  болады,  мәселен,  «…екен  деп 

болжам жасауға болады».

көрсететін  мысалдар  да  келтірдім.  Мен,  басынан  бастап,  сұхбат  жүргізу 

барысында, зерттеуші орталық тақырыптар мен зерттеу мәселелеріне белгілі 

бір онлайн талдау (кейде жауап алушымен бірге) жүргізе алатынын айттым. 

Келесі  кезең  –  транскрипциялау  жүйесін  жүйелі  түрде  қолдану  есебінен 

жазылым  үлгісіне  бастапқы  мәліметтерді  айналдыру  болып  табылады. 

Индуктивті  және  дедуктивті  тәсілдердің  қоспасының  есебінен  түзілуі 

мүмкін кодтау ұстанымдары арқылы кейіннен жазылымды кодтауға болады. 

Ұстанымдар әрбір көрсетілген кодтың сипаттамасын қамтуы және есептің 

әдістемелік тарауында негізделген болуы тиіс. Код – бұл белгілі бір зерттеу 

концепциясына қатысты затбелгі болып табылады. Ашық кодтау жазылымды 

түрлі концепциялармен байланысты үзінділерге бөледі. Тақырыптық кодтау 

түрлі  ашық  кодтар  арасындағы  байланыс  негізінде  теориялық  талдаудың 

мейлінше  жоғарғы  деңгейіне  жеткізеді.  Компьютерлік  сапалық  талдаудың 

артықшылығы – ол мәліметтерді сақтау, ұйымдастыру және өңдеу үдерісін 

жеңілдетеді. Визуалдандыру мен матрицалар екі негізгі себептер бойынша 

сапалық  мәліметтерді  талдаудың  пайдалы  құралы  болып  табылады:  олар 

талдауды құрылымдау үшін база ретінде қызмет етеді; және ұзақ мерзімдік 

зерттеу кезінде жиі қолданған жағдайда талдау барысы туралы ақпаратты 

қамтиды. Визуалдандыру мен компьютерлік бағдарлама көмегімен түзілген 

матрицалар бастапқы кодталған жазылымға гиперсілтеме ұсынады. 

Analysing Qualitative Data

423

Қосымша оқу үшін 

Saldana, J. (2009) The Coding Manual for Qualitative Researchers. London: Sage.

Silverman, D. (2010) Interpreting Qualitative Data, (3rd edn). London: Sage.

Taking A Quantitative Approach

424

Taking A Quantitative Approach

425

TAKING A QUANTITATIVE APPROACH

MARK WINTERBOTTOM

CHAPTER 9

Chapter Overview

A quantitative approach means using measurements and numbers to help formulate 

and test ideas. It usually involves summarizing numerical data and/or using them to 

look for differences and associations between sets of numbers. In this chapter, I’ll look 

at approaches to collecting and interpreting quantitative data. In the next chapter, 

you’ll learn more about using statistics to analyse them.

If you have a natural science background, you may feel at home here, but bear in 

mind the complexity of human behaviour – don’t ignore the depth of data available 

through the qualitative approaches outlined elsewhere in this book – achieving a fully 

natural scientific approach in a school context is almost impossible. Read this chapter 

together with Chapter 12. Never collect a set of data before thinking about how to 

analyse it!

Before I get going, let’s look at some fundamental words and ideas, which can help 

you to talk about, read about, evaluate and plan quantitative approaches. I’ll then 

introduce you to two approaches for planning your own research, and look at some 

ways in which school performance data is used ... and misused!

Taking A Quantitative Approach

426

IDEAS AND DEFINITIONS

Variables

A quantitative approach usually means measuring a property of something or someone.

That property is called a variable. Variables are called variables because they are

entities that can vary. You can collect quantitative data about individuals by designing

questionnaires or tests. Alternatively, you can simply record data by observing the subjects

‘from afar’; it all depends on the data you want. However, do bear in mind that

the act of collecting data can sometimes change the data you get! Some examples of

variables include:

•  the number of students ‘on roll’

•  the test result

•  the proportion of students gaining five GCSEs at A–C

•  the tier (e.g. primary, secondary, etc.)

•  the school governance system.

Variables described using numbers are quantitative (e.g. the proportion of students

gaining five GCSEs at A–C). Those described by categories are qualitative or categorical

(e.g. the school governance system – foundation, voluntary-aided, etc.). Although this

chapter is about a quantitative approach, we usually look at qualitative variables as well.

Quantitative variables fall into two types. Continuous variables can take any value in a

given range (e.g. 3.2, 4.798), whereas discrete variables have clear steps between their

possible values (for example, you can’t get 100.324 pupils at a school!).

Another way to think about variables is the scale, or level of measurement that

we use. The scale itself determines whether they’ll be qualitative/quantitative or

continuous/discrete.

•  Nominal scales are for qualitative variables to categorize observations. The value assigned 

to a group is just a label, and implies nothing about quantity. Sex would be a variable 

measured with a nominal scale: we could use ‘1’ for boys and ‘2’ for girls.

•  Ordinal scales assess rank or order and yield discrete variables. Imagine you rank the 

pupils in your class according to test score; the best student has a rank of 1, etc. Rather 

than just being labels (as above), a ‘1’ is better than a ‘2’. This type of scale gives an 

effective summary, but the exact value of the differences between each person’s scores 

is unclear, and not necessarily identical.

•  Interval and ratio scales provide discrete or continuous data where there are equal 

intervals between the units of measurement (for example, a minute is the same length 

however you measure it!). In a ratio scale, a zero means zero – there are no children or 

they got no answers right. In an interval scale, the zero is relative – e.g. zero may describe 

a baseline motivation level.

Taking A Quantitative Approach

427

Validity and reliability

Some variables are direct measures of what we’re interested in – recording a pupil’s sex is a 

direct measure of their sex. Others are more indirect measures. For example, test scores are 

an indirect measure of pupils’ learning. It is important that such measures provide a genuine 

measure of the underlying construct. Such validity is important when collecting your own 

data, and when interpreting other people’s data and conclusions.

It is also important to consider reliability – the consistency and repeatability of data collected 

over time, across different samples, and across different measures of the same underlying 

construct. Box 11.1 suggests some ways to assess reliability using one pilot group, or using 

two groups that are closely matched for variables relevant to your study.

Samples and populations

When collecting quantitative data, distinguishing between your sample and your population 

is important. A sample is a sub-group of the population. Collecting data about a sample that 

is representative of a wider population lets you draw conclusions about the population. We 

often use samples because measuring all the individuals in the population is impractical.

To ensure your sample is representative, it’s essential to understand who your population 

is – this is something it is easy to overlook. For example, if you want a study which is 

generalizable to the population of all the 14-year-old students in the country, then you 

would randomly choose your sample from all the 14-year-olds in the country – this is so-

called probability sampling. You can see four types of probability sampling in Box 11.2.

BOX 9.1 

Assessing reliability

• Use one questionnaire with one group on different occasions and see if their 

answers are significantly correlated (see Chapter 12). (Be aware that they may 

remember their responses though!).

• Use two different questionnaires (but whose questions examine the same 

ideas) on different occasions with the same group, and look at the consistency 

between responses to matched questions.

• Use one questionnaire with both groups, and check that their responses are 

not significantly different.

•  Use  two  questionnaires  (whose  questions  examine  the  same  ideas)  on 

different occasions with both groups and check that responses to matched 

questions are not significantly different.

Taking A Quantitative Approach

428

However, your own research is likely to happen within your own school and often in

your own classroom – this is so-called convenience sampling, a type of non-probability

sampling (see Box 11.3). The children in your class are not representative of all the

14-year-old children across the country, and you cannot therefore make generalizations.

Your pupils’ characteristics may be dependent on upbringing, socio-economic group,

location, year group, on your idiosyncrasies as a teacher, and many other variables.

Hence, when conducting research in your own classroom, your class is the population –

there isn’t a wider group to which you can generalize your findings.

BOX 9.2 

Probability sampling

• Random sampling: Here there is an equal chance that each member of the

population is included. Including one individual in the sample has no influence

on whether another individual is included.

• Systematic sampling: Sometimes practicalities may make it preferable to

sample individuals in some sort of order – say every fourth subject in a line.

To do so, you should randomize the list of individuals and choose your starting

point randomly.

• Stratified sampling: You may suspect that other variables (e.g. sex) could affect

your results. To try to eliminate the effect, you would randomly choose half your

subjects from the boys and half your subjects from the girls.

•  Stage  sampling: You  can  stratify  your  sampling  at  a  number  of  levels.  For 

example,

if you thought that the year group and tutor could affect your data, you

would randomly choose year group, then within each, randomly choose tutor

groups, and then within each again, randomly choose the pupils to study.

BOX 9.2 

Non-probability sampling

• Convenience sampling: You use individuals to whom you have easy access. You

cannot generalize your conclusions to a wider population.

• Quota sampling: You may suspect a particular variable (e.g. sex) could affect

your results. To eliminate any influence, randomly choose your subjects from

boys and girls, but in proportion to the number of boys and girls in the group

of pupils you are interested in (e.g. the population of pupils in Cambridgeshire).

Your findings are only generalizable to this limited population of pupils.

Taking A Quantitative Approach

429

If you have a natural science background, you may feel this makes your research rather 

pointless – after all, if it’s got no wider application, what’s the point in doing it? There are 

two answers: (1) researching your own practice in your own classroom contributes strong-

ly to your ongoing professional development; and (2) providing you make the context of 

your research clear when you write it up (you, the nature of the class, the lesson content, 

the whole school context, etc.), anyone reading your study would be able to decide the 

extent to which your findings may apply to them (so-called user generalizability).

Finally, although you probably won’t need to generalize to wider populations in your own 

research, you will read large-scale studies that do just that. Even if individuals have been 

sampled randomly, instinct probably tells you that a study based on two individuals is less 

generalizable than one based on two hundred – but why is that?

Well, if you take lots of different samples from the same population, it’s unlikely that each 

will have the same mean (what most people would call the average) or standard deviation 

(how much the data is spread out around the mean) for the variable you’re measuring. 

However, if you use a bigger sample, the mean will be closer to the population mean; 

hence, the bigger the sample, the better. If you do adopt a quantitative approach, a sample 

size of 30 or more would be good.

QUANTITATIVE APPROACHES TO RESEARCH 

So how do you actually generate some data? There are two key approaches: experimen-

tal (measuring the effect of some sort of intervention), and non-experimental (looking at 

what’s there and trying to make sense of how different variables may affect each other). 

You’ll learn how to analyse your data in the next chapter. 

Experimental

This approach looks at the effect of one variable on another, by making a change in one of 

the variables (the independent variable) and seeing how the other variable changes (the 

dependent variable), while keeping all other variables constant (controlling

them). An experimental approach is broadly underpinned by the stages shown in Box 9.4.

• Dimensional sampling: If you suspect that a number of variables will influence 

the variable you are interested in, you can deliberately choose individuals who 

are subject to every combination of those variables.

• Purposive sampling: You choose which individuals will be in your sample based 

on how representative you think they are of the group you want to study. It 

is unwise to generalize beyond your sample as your choices are unlikely to be 

fully objective.

Taking A Quantitative Approach

430

BOX 9.4 

Planning an experiment

1.  Identify what you’re trying to find out, and work out what you think will 

happen.  Predict  how  you  are  expecting  the  independent  variable  (the 

treatment) to affect the dependent variable (the response).

2.  A variable like ‘pupils’ learning’ cannot be measured directly, and you’ll 

have to use a ‘proxy’ or indirect measure of it, such as ‘test score results’; 

remember to justify your choice of ‘proxy’ as a valid indicator of the variable 

you’re interested in.

3.  Decide who your population is, and then randomly choose a sample of 

individuals from the population.

4.  If  your  experimental ‘treatment’  is  something  like ‘receives  new  teaching 

approach’ or ‘doesn’t receive new teaching approach’, then randomly 

allocate your pupils to each group. Even if your experiment involves a more 

quantitative independent variable, such as the length of time spent working 

on computers during a lesson, randomly allocating pupils evenly across 

levels (one hour, two hours, etc.) is still essential.

5.  Make sure that the levels of treatment are realistic within the context 

(usually a classroom). For example, looking at how eight hours of computer 

access affects pupils’ learning is unrealistic in a single lesson. Also ensure 

that the range of treatments you provide will enable you to see trends and 

differences. Comparing 60 minutes of computer access with 61 minutes 

won’t yield any startling conclusions.

6.  Identify potentially confounding variables (variables that could affect your 

findings, such as variable C in Figure 11.2), and develop strategies to control 

them (keep their levels constant between individuals receiving the different 

levels of treatment), or eliminate them (e.g. removing a teaching assistant 

from the room).

7.  Pilot your methods with a different sample to iron out any difficulties.

8.  After the experiment, measure the dependent variable for each individual 

and use a statistical test to make a conclusion. Don’t leave it until now 

though to consider which statistical tests you intend to use – it is all too 

common to realize that the data you’ve collected is not compatible with 

any statistical test.

9.  Be careful to state the extent to which you can generalize your conclusion 

across a wider population.

Taking A Quantitative Approach

431

An experimental model is a good basis for scientific research in a laboratory, but employ-

ing this approach in a classroom, with the multiplicity of variables in existence, and the 

constraints of school timetabling, is not always easy.

Let’s imagine you’re researching the effect of a six-week motivational training programme 

with a group of Year 9 students. Rather than being constrained by the groups already set 

up, you’ve randomly selected a group of students and you’re teaching them at lunchtime. 

You use a questionnaire to assess their motivation beforehand. You use the same ques-

tionnaire after the six-week programme and find an increase in motivational levels. To your 

delight, you conclude that the training programme has had a positive impact!

Or do you? Are you certain that you’ve controlled all other variables that may affect your 

results? You may have begun your programme at the start of term, and then measured 

motivation again at the end of term; perhaps you could expect children to be more moti-

vated as the end of term draws near! In fact, all sorts of other variables could be responsi-

ble for your findings. Hence, thinking about and collecting additional data to examine such 

confounding variables is essential to working out what’s really going on. If you do want to 

use just one group, you may find your work sits better as one cycle of an action research 

approach, particularly if you were unable to sample children randomly.

Using a control group

However, if we stick with an experimental approach, how can we get rid of the effect of 

these extra variables? The easiest way is to use more than one group whose members 

have been chosen randomly. One group gets the training programme and one group 

doesn’t.

Because you’ve chosen your groups randomly, any systematic effect of the other variables 

should be spread across the groups, and will ‘confound’ your results to the same extent. 

You therefore look at the effect of the training programme by comparing the increase 

in motivation for the ‘trained’ group against the increase for the ‘untrained’ group. Even 

though both groups may have greater motivation at the end of term, any additional effect 

of the training programme should be clearer.

It’s still not necessarily simple though.

•  Just  doing  something  with  your  experimental  group  (even  if  you  gave  them  a  free 

lunch  for  six  weeks)  may  affect  their  motivation.  Hence,  leaving  your  control  group 

with no intervention at all would not be appropriate – you’ll have to consider what an 

appropriate control treatment would be.

•  You need to be careful of the effects of other variables ‘creeping in’. For example, if you 

taught one group on Monday lunchtime and the other on Tuesday lunchtime, you’re 

introducing another variable which could bias your outcomes.

•  Variables can interact in unpredictable ways. For example, being given a pre-test may 

actually influence the results of the post-test (children may think about their responses 

between the two tests and want to ‘put the right thing’), particularly if the tests are 

very  similar.  Comparing  your  results  with  two  further  groups  (one  control  and  one 

experimental) which do not experience the pre-test would help to clarify the extent of 

this problem.

• 

Taking A Quantitative Approach

432

Identifying these problems, and collecting some supplementary data to explore them,

will add weight to your study when you write it up. 

Quasi-experiments

So what happens if you cannot choose groups randomly, you cannot control variables,

etc. Well, you have to do the best you can. A ‘quasi-experiment’ is probably as good as

you’re going to get.

This approach still uses an experimental and control group, but rather than being able

to randomly choose the members of each, you should choose existing groups (to which

you have access) which are most similar on as many relevant variables as possible. This

means that if you’re looking at motivation levels, you really want two groups that have

very similar motivation levels in the first place, and a similar range of other variables that

could be important, such as prior attainment and socio-economic factors.

Rather than just looking at the overall differences between the two groups, comparing

the effect of the experiment between pairs of pupils from the experimental and control

group, matched as closely as possible on such relevant variables, is even better.  

жүктеу/скачать 32,4 Mb.

Достарыңызбен бөлісу: