I've been wading through this book and thought I'd share some notes with you. This is how I have understood things, so if you feel that there is something you do not agree with. Please feel free to comment. Educational research procedures are both public and systematic. Educational research itself needs to be purposeful, should also be useful and is normally conducted in classrooms, labs, libraries and informal learning situations. It is governed by rules which specify using correct methods, explain
how one has done their research, inform readers of every important point found and which treat all participants in an ethical fashion.
Novice researchers also need to be aware of conventions which detail the ways in which they have traditionally approached an area or topic. They also need to note the values and limits of these conventions.
Educational Research contributes to shared knowledge and researchers assume that their readers possess an information base made up of both GLOBAL and SPECIFIC shared knowledge.
Article content matters; it must not be deliberately false, should strive to create something new and be accessible to its target readers. Educational research articles can be tackled part by part so start with what you know.
Secondary articles in the form of lay reviews, action plans, interviews or opinion pieces can be used to understand primary articles. However, if one feels that information in these is too good to be true or oversimplified go back to the original and check it out for yourself. Mastering the primary article, taking it apart and engaging with it is far more effective than relying on a secondary article to do this for you; secondary articles should be used as tools in aiding understanding.
QUANTITATIVE LITERACY
Personally, this area scares me a little and I think this has its roots in secondary education mathematics (but hey I'll give it a go anyway?!).
What about variables? A variable is a quantity / element / feature or factor that is liable to variance or change. In scientific research, variables are divided into the dependent and the independent. The former are target variables (something we want to change) while the latter have an impact on the former. An independent variable is one whose value does not depend on that of another (x) while a dependent variable is a variable which does depends on that of another (y).
Types of Articles:
Scientific research is often presented through different types of article. Here are four types:
1. Descriptive articles look at describing processes or situations. This focuses on studying how a poorly understood process or situation might operate in a natural, uncontrolled environment. When looking at opinions or beliefs for example, we often measure them in terms of likelihood or frequency. With this in mind, we are then able to turn to descriptive methods to understand how this information can be better understood, compared or contrasted to other groups.
2. Relation-finding articles are written by researchers interested in establishing that there are perhaps relationships between two variables. The move from a descriptive article to a relation-finding article denotes a move from searching for patterns to a search for impacts.
3. Hypothesis-testing articles bring in the use of (surprise, surprise) hypotheses to say powerful things about how independent and dependent variables are related.
4. Model-building articles move away from descriptive articles looking at what patterns variables reveal to building models which will allow us to put them to work. These models are usually based around a target variable which is of particular interest to researchers. A model is developed when the pattern of other variables are used to say something about the target variable.
The Scientific Method
I understand the scientific method as a cycle of refining knowledge. It begins with a hypothesis or prediction, we then collect data and see if they match with the hypothesis. If the answer is yes, then we are on the right path. If the answer is no then we need to change the hypothesis. It is therefore a cycle of prediction, test, change and so on and so forth. The scientific method is also public, replicable, fallible and correctable, generalisable and is able to simplify understanding of complex things.
A Statistical Worldview
Science introduced the idea of a Newtonian deterministic model; one that says that if you know all your causes, and all of our beginning conditions, then you can predict and describe all your efforts precisely and exactly. However, eventually the sciences began to abandon deterministic models and replace them with probabilistic; a statistical worldview that recognises that probability is built into the very fabric of reality and within the framework of probability, there are systems of order and also stability in the world.
A Quantitative Mini Glossary (Non-Alphabetised)
At this point I thought it might be a good idea to look at some of the language used in scientific articles. Here is a mini glossary:
A datum - An individual score or measurement
Data - A collection of scores or measurements
Distribution - A collection of measurements for a single event or category
Constant Distribution - When all the members of a distribution have the same value
Blob Distribution - When there are no structures or orderly patterns to discover (and you are on the wrong track).
Normal Distribution - When there is a meaningful typical score for the distribution and an orderly process of change that accounts for differences between this typical score and other given scores.
Central Tendency - Clustering in a distribution. There are three common types (Commonness, Clustering above / below and Average Score - See below).
Commonness - This focuses on the mode; the most common. This divides distribution into two lists, the mode score and the rest. In order to be the mode, it merely has to have a higher frequency and it does not matter how much of a higher frequency of occurrence it has. When there is more than one score that has the highest frequency, we refer to this as multi-modal distribution.
Clustering Above/Below - This focuses on the median. An example - Let's take a group of trees, measure their height, and display them in order of their height. When you hit the middle, you have hit the median. This is useful when looking at ranking.
Average Score - This focuses on the mean and is based on the value of ever item in the distribution. This changes if one score changes in the distribution. It is the balance point of the distribution and is the first best guess of any given score in a distribution.
Dispersion* - When there is more than one score and the distribution is spread out in some fashion. This property of distribution to spread out is called Dispersion.
On that note here is a type of dispersion: Standard Deviation
Standard Deviation
I cringed at this at first... but here goes...
Apparently, standard deviation 'captures' the average distance of any given score from the mean. Because the mean is balanced exactly between anything below it and anything above it - any differences cancel themselves out. So the differences are converted from looking at them from a 'above/below' perspective to an 'away from the mean' perspective. I am sure that Shank and Brown had the very best of intentions regarding me understanding this and all I can say is that I was probably in a coma at school when my maths teacher delved into the subject!... But I think I got it in the end.
The size of Standard Deviation (SD) is a good indicator to the degree of variability. In other words, if the SD is large then there are less scores near the mean. If it is small then vice versa. When moving away from the mean, one standard deviation at a time, not only are the events becoming less likely, but the processes behind creating those events are less like the processes that shaped the scores around the mean.
Distributions from Populations or Samples
I have discovered (or I think I have) that distribution is governed by parameters; mean and standard deviation values. But where does this data come from? The answer is a sample. A sample is a subset of a population and sample data are from distributions with means and SDs - a.k.a. statistics... there, I said it and got it out of my system. Now, there are several types of samples you can have:
A convenience sample where the members of a population happen to be conveniently close at hand e.g. work colleagues, your class of five-year-olds, though this is considered the least acceptable type of sample unless you can convince your readers that members of a nearby sample are no different from your own.
A purposive sample where members are chosen because of specific characteristics. In this case you treat the sample as if it were a mini population. There is a restriction in this because there is less generalisation possible so its better for more qualitative research.
A representative sample (or stratified sample) is a population in miniature.
But the crème de la crème for scientific researchers is... the random sample! The random sample has members drawn at random from a population. For scientific researchers, this is the sampling strategy winner.
Stats, Stats and more Stats!
I have to admit that at this point I started to freak out a bit. I was out of my 'qualitative comfort zone' but I'll continue. Scientific researchers use statistics to test hypotheses and assumptions they have about the world and this is known as inferential statistics. At this point I'm going to throw around with a few words.
Correlation - the correlation coefficient is used to determine the probability that two variables are related. The most common is known as the Pearson Product.
A T-test - comparison between two means. A target variable may be under two different conditions. Values from a -test allow researchers to decide whether or not to include these items in a final equation or not.
ANOVA - from ANalysis Of VArience. This does what a T-test does but allows us to compare more than two groups.
Chi-Square - Question: Do the frequencies we observe in the world match those we might expect, or is something different going on? Chi-square helps to answer this.
I suppose that if you want to go down this route (I am not convinced I will) you can pick up a few good stats books to help you along the way.
The authors then get into talking about multiple regression equations, path analysis models, factor analysis and structural equation models at which point I my head exploded so I drifted off into academic purgatory. Please feel free to clarify and comment by the way. I am venturing into safer territory (I think) and looking a Qualitative Educational Research so watch this space!
* ACCORDING TO THE AUTHORS, YOU NEED TO LOOK AT BOTH CENTRAL TENDENCY AND DISPERSION TOGETHER BECAUSE DIFFERENT VALUES ARE ORDERLY IN A PROBABILISTIC WAY AND YOU NEED BOTH TO MAKE SENSE OF THE SCORES.
