Tuesday, December 23, 2008

9 Steps to Research Process - 1.Question?

The remaining post will be an introduction to how research takes place. It seems appropriate in understanding research as the discriminatory process that we train to understand anything external or internal.

How People Do Research: simplified representation of the research process involving 9 steps: identify a question, looking at what others have done, refining the question, developing a background, choosing a method, preparing a proposal, collecting the data, analysing the data, reporting the study. Although we will go through these steps sequentially, there is often side-stepping, back-stepping, and skipping of steps. In order of steps, asking a question is the first step.

1. IDENTIFY A QUESTION

Is the question a researchable question?

A question must have a theory base, be significant, have appropriate variables, require appropriate resources and have available subjects to be researchable

Theory Base
If there is some knowledge about the area is is often easier to research. You can see what others have done and add to what they have started. If there is a theory covering the question you wish to ask, so much the better. A theory is a way of organising the knowledge into simple principles and relationships. A theory should help you to understand the area better, and make sure you don;t miss things you should have thought of.

Significant
Research takes a lot of time, thought and effort. It is something you should be proud of. Your research should be interesting to some people, and be worthwhile to some people. If everyone you discuss your research question with says, "So what?", perhasp the question isn't significant.

Variables
A variable is an attribute or characteristic that can vary. Whether your question is researchable depends on the variables you intend to measure. Variables range from being easy to identify to being hard to describe, let alone identify. Some variables are fairly easy to measure whilst others are very difficult. To be researchable your question must involve variables which are measureable in a reliable and valid way. The variable(s) you measure is called the dependent variable (DV) and the variable(s) you control is called the independent variable (IV). Remember, the DV in one study may be the IV of another study.

Resources
Your research question needs to be logistically feasible in the resources it requires. Resources include money, time, equipment, space and technical assistance.

Subjects
Depending on what characteristics your subjects must have, and how many of them you need, your subjects may be easy or impossible to find.

Types of Research Questions


When you have a potentially researchable question, you need to decide what type of question it is. There are 3 main types of questions, corresponding to the three types of research:

Descriptive Question: What is A like?
Correlation Question: Is A related to B?
Experimental Question: Is A different to B?

To be able to answer any of these questions, the variables, the As and Bs of research questions, need to be measureable in a reliable and valid way.

Reliability

Reliability is synonymous with repeatability, stability and consistency. It concerns whether in the same situation, a test will give the same results on different occassions or, whether different parts of a test measuring the same phenomena will give the same results.

In any measurement there will be the true value plus or minus some error. This error can be systematic or random. Systematic error is inaccuracy. Random error is unreliability. When the measurement is accurate and reliable there is minimal error.

There are 3 main sources of unreliability; tester error, instument error and DV variability.

There are 4 types of reliability;

i) intra-tester reliablility: the degree to which the same tester can get the same results at different applications of the test, that is whether the test is stable over time.

ii) inter-tester reliablilty: the degree to which different testers can get the same result when using the same test in the same situation, whether concurrently or sequentially

iii) parallels or alternate forms of reliability: the degree to which two versions of the same test can give the same result

iv) internal consistency: the degree to which two questions covering the same content in a test can give the same results, that is like parallel forms but within the one test.

You assess whether the variables your question involves are reliably measureable by looking at evidence from literature or similar situations and/or evidence from a pilot study. It is directly testable. Use an inferential statistic to determine whether there is any difference between testers or sessions.

Validity

In contrast to reliability, validity is not directly testable, and is not easy to fully define. 4 types of validity (for experimental questions) are:

i) Statistical conclusion validity. This type of validity concerns whether there was a real difference in the data, or whether the observed difference was jsut due to "chance". This is mathematically testable. (law of higher probabilities to statistical significance)

ii) Internal validity. This concerns whether the difference in the DV was caused by the IV. This type of validity has to be argued by showing that only the IV could have resulted in the change in the DV, that is the change in the DV couldn't be realistically expected to have been caused by something else.

iii) Construct validity concerns whether the DV adn IV really represent the constructs they are supposed to. That is, have you been modifying and measuring the thing you say you have.

iv) External validity concerns whether the results you found in your study are likely to be found with other people, in other places and in other times.

If your question is a correlation question, all of these types of validity apply equally, with statistical validity saying whether the association/relation you found was likely to be due to "chance" or not.

If your question is descriptive, statistical, construct and external validity apply.

Except for statistical validity, validity has to be argued. You need to be able to convince any reader that you are researching what you say you are. to provide evidence to convince the reader of construct validity, you can show that it looks as if it is what you say (face validity), and that it measures the same as a respected measure (criterion-referenced validity)

Conclusion

The first step in starting research is finding a researchable question. Where the questions come from varies, but whatever it came from it should create a desire to discover the answer. And Data that is not reliable or valid is not real data.

No comments: