Deck 4
Main point of intro:
operationalization (the operations by which you transform constructs in your RQs into a measurement plan)
operationalization cuts through the boundaries between conceptual and technical design it starts out with specifying a logic of classification (how and why do we conceptually
express this construct as a combination of aspects or indicators
then proceeds to tie a technical plan of instrumentation and measurement procedures to each indicator.
When you follow your operationalization and collect data on the specified indicators, you are trying to find out more about the characteristics of some things or beings. Which are
those units? What is the target population you want to make inferences about? Does this population coincide with the population you extract your data from, or will you, say, use
bioindicators (e.g. a certain kind of fish) for your measurements so you can make inferences about a characteristic of your target population (pollution of the lake)?
Operationalization and population support each other (you need to decide on them in tandem if you want things to work out well). For population, the link to technical design is
sampling: you can’t think about which units from your population you sample if you haven’t delineated which population(s) you’re interested in!
Operationalization
Operationalization specifying (and justifying) the operations needed to measure key constructs. All scientific research contains constructs that need to be operationalized.
--- Temperature as ‘ most likely distribution of molecules over energy states in a system at thermal equilibrium’ is only proxied by the actual measurements like expansion of mercury. In some
cases these proxies are well-understood and solid, but in other cases we need to make stronger assumptions. This is often the case in surveys, where we assume that our questions are interpreted
and cognitive processed in certain ways, that individuals have attitudes that they know and that they can and will express in certain ways when asked about it.
The operationalization can be quite simple (take the degrees Celsius on a thermometer), or become very complicated (biodiversity).
Marital status is a reasonably simple construct. This can be turned into a number by using a questionnaire in which you specify categories of marital status and then respondents have to indicate
to which category they belong. However, you still have to think about the relevant categories for your research (do you want to distinguish between widowers, divorced?)
Temperature may be complicated to measure technically (how do I measure temperatures in a big lake at various depths? If I use remote sensing, how do I account for permittivity and
atmospheric scattering? Etc.), but conceptually it’s a simple concept that doesn’t typically need to be decomposed into indicators using a tree diagram.
Biodiversity is much more complex. Here, you do need to think about dimensions and indicators. Directly trying to measure biodiversity with one or two variables is problematic. Number of
species doesn’t cover biodiversity. If I have 10 species, each with 100 individuals, few people would rank that as equally biodiverse as an area with 10 species, 1 of them with 990 individuals
and 9 species with 1 individual each. This means we’ll also need to think about diversity. And what if two areas have equal numbers and evenness of species, but in one area all species are
habitat generalists with a limited set of ecological functions and in the other there is a mix of specialists and generalists with a variety of functions. Should we also take that into account?
The last example, about a specific kind of attitude towards a specific kind of object, seems so specific that it couldn’t possibly be hard to operationalize, but it’s in fact still a complex construct,
because the feelings about this object could still consider various aspects of the object (design, texture, feel to lips, grip etc.)
Key steps of operationalization
Constructs have aspects = indicators are expressed on a clear measurement scale (variables & are measured with certain instruments and protocols for using those instruments.
Everything up to indicators is the conceptual part of operationalization. The rest is technical.
Example 1 - GRQ: What is the effectiveness of
pulverized drumstick tree seeds compared to other (chemical) methods of clearing water?
Step 1: construct to be operationalized
Effectiveness is a complex construct and we’ll need to think about its aspects.
,Step 2: indicators
Here we do so using a simple tree diagram. Effectiveness is sometimes just understood to be the degree to which it has an effect on the
relevant outcome (in this case: how well it clears the water). However, often effectiveness is used more broadly to include cost-
effectiveness, time-effectiveness/efficiency or even user-friendliness.
A better version. Often, the way we approach something like ‘effectiveness’ is depend on what we are studying (effectiveness of what?).
Same with attitudes towards [something]. In these cases, instead of picking ‘effectiveness’ or ‘attitude’ as key construct for
operationalization, it’s better to pick ‘effectiveness of [thing]’ or ‘attitude towards [thing]’ as construct. In your aspects, try to be clear what
you mean. What is time? Time it takes for the substance to clear the water, time it needs to gather the seeds, crush them, apply them and
clear the water? By labelling it as either ‘reaction time’ or ‘process time’ it’s already more clear. Same for the other categories.
What do you think about narrowing costs to financial costs?
How about amount of active substance? Can we simply compare 1 gr of a strongly concentrated light-weight chemical to 1 gr of diluted heavy biological treatment?
Step 3: Variables ndicators into variables by adding a measurement unit or scale.
Example 2 - To what extent is there a relationship between bee prosperity and habitat type?
Step 1: prospering of wild bees
Step 2: What are good indicators for prosperity of bee? Pop size+diversity. How about health of the bees? And, as we’ve seen, ‘ diversity’ is still too broad as an indicator.
Step 3:
Bee traps: colourful, low buckets with water and soap. Bees are attracted to colour, get stuck in trap. Above: network of bee traps.
Put in grassy habitat and bushy habitat for three days in July. Then number of bee (species) in traps are counted and compared.
Any issues?
• If bushy and grassy habitats attract different bee species with different sensitivities to color or differences in ‘ boldness’ , different rates of capture may occur
• Better at trapping small insects which means bias towards certain species, young insects etc.
• Increased capture expected in grassy areas because of fewer competing visual cues
• Effectiveness of traps depends on elements and if wind blows more freely in grass than in bush habitats, this may be an issue
If you plan on repeating measurements over time, it’s also good to think about other issues, such as whether measuring affects the thing I want to measure. If I disturb the bee population or kill
many bees my capturing them, that may result in fewer bees in my next measurement
Measurement of scales
Simply describing the research instrument is not enough. How do we use it? Operationalization also means going into the actual operations for using the research instruments.
• Nominal is the most simple one. These are just labels, like gender or type of specie. There is no rank order.
• Ordinal also has labels, but they are ranked in a meaningful way. Like the top 10 of the most livable cities, or educational level.
• There is, however, no equal distance between the categories.
• An interval scale is rank-ordered with equal steps. For instance the difference between 18 and 19 degrees celcius is the same as between 31 and 32.
• But there is no natural zero, this means that zero degrees celcius still is a temperature, which is arbitrarily chosen. Zero degrees celcius equals 32 degrees Fahrenheit. (Kelvin does
have a natural zero, the temperature at which all thermal motion ceases)
• Ratio is the easiest scale, because it represents numbers as we usually think of them.
• Ratio means that there is a rank-order, that there are equal distances between the values and that there is a natural zero.
• Zero means absence of a variable, like the number of children.
, The most important reason that we distinguish between the scales, is that it determines the statistical procedures you can use on them when describing and analyzing your data.
Note: ‘higher’ measurement levels are not necessarily better. To measure marital status, a nominal scale is as good as it gets.
Also note: concepts do not (always) have a ‘natural’ best measurement level. Income is not naturally a certain kind of variable already (we mentioned ‘ Kumar’ here as a writer who claims this;
no worries, this refers to a textbook we once used, but which you no longer have to read...lucky you!). Can you think of ways to measure income with different measurement scales we
discussed? [Main source of income: 1. work income 2. rent 3. profit from business 4. stocks etc. = nominal, How big would you say your income is compared to others? 1. Below average 2.
Average 3. Above average = ordinal, How big is your annual income from work? .... Euros = ratio.]
• So, all measurement scales distinguish between values.
• All but nominal have an ordering of values.
• Interval and ratio in addition have equal distance between the values.
• And ratio has a natural zero, which means that you can calculate with it.
• In some cases, ordinal can have a natural zero. For instance with income in non-equidistant categories.
Another example:
Color can be nominal: color names
Ordinal: unequal intervals of wavelength
Interval: angle in the hue color palette
Ratio: wavelength (but is 0 nm wavelength really a natural 0?)
Population
Research pop VS research pop unit
Target pop VS pop is not always the same as the pop want to extract data from
- RQ=starting point, sometimes pop already mentioned or directly implied there
- If not, be extra careful: you need to clearly delineate your population of interest. You need to optimize your research design to make inference to a particular population, so you'd better be
clear about which population (in context and time domain)
In the operational population, you come to a working definition of the terms in your theoretical population (mature=? Bottomfeeding organisms=? 'close'=?)
Continuous populations pose more problems for discretization into units, but it is important to consider the dimensionality of units. This will also have implications for your sampling design.
One particular concern is expected heterogeneity (if you defined huge spatial units but you expect big variation every 5 cm, wouldn't it have been better to choose smaller units?).
Sometimes there's just your population and that's it. I want to make inferences about fish and to do so, I'll measure stuff on fish. Often, however, things are more complicated...For any relevant
population, you need to go through the same process of providing a theoretical and operational definition and indicating the discretization into meaningful units. In this particular example, the
research units will be characterized by measuring observation units from a single population, but more complex situations may require multiple populations (e.g. we could also study sediment
and water in order to characterize the same research units).
You typically have a difference between research and observation units when you need or want an indicator/proxy to characterize the research units (e.g. bio-indicator) or you use nesting in
hierarchical systems (e.g. aggregate over measurements on trees to make inferences about characteristics of forests).
Population under certain conditions, during a certain season etc.
When sampling, you will often select research and/or observation units, but there may also be separate sampling units. For example, we may sample monitoring stations to measure air pollution
in an area, even though these stations are neither research nor observation units. We may also sample seats in a theatre rather than the persons sitting on them as a way of getting to our actual
observation units (the people whose behinds keep the seats warm).