Null and research hypotheses: distinguishing directional from non-directional approaches

¶ … distinguishes a null hypothesis from a research hypothesis? How would you explain the difference between non- directional research and directional research hypothesis?

A good research hypothesis is based in previous research, and is constructed in an experimental form so it can be reproduced by other researchers. A formal statement of the research question is called a hypothesis and is phrased in a 'yes or no' or 'true and false' fashion. A good research hypothesis usually allows for a control group. Unlike a null hypothesis, a research hypothesis can be tested by actively observed, obtained, or tested research, while a null hypothesis merely explains a body of data. For example, a null hypothesis might be that girls who are educated in female-only classrooms are more likely to do better on their SATs than girls who go to schools with coeducational mathematics classrooms, after examining data on the SATs and high school student demographics.

Directed research actively creates an experiment to test a particular hypothesis. For example, directed research might subject a group of kindergarteners to a specific type of enrichment preschool program and use students with comparable socio-demographic characteristics as the control group. Non-directed research studies study individuals in the field over time, but the researcher does not specifically direct the research to conform to a hypothesis. For example a non-directed research program might study students who participate in Head Start as preschoolers and those students who do not and determine if participation in Head Start seems to correlate to a certain level performance, positive or negative.

Unit 2

Question

What is the difference between reliability and validity? Discuss the importance of reliability to the research process. What is the danger in using an unreliable measure when testing a research hypothesis?

Reliability refers to how consistently a result may be produced when performing an experiment again by other researcher. Reliability directly relates to whether an experiment can be said to validly prove a hypothesis, although using a reliable test instrument is not the same thing as proving a hypothesis. If a particular experimental technique or instrument can measure and produce the same results through a series of trials it is likely to be reliable, and if the research process yields the same result when done by different experimenters later on, the research instrument is even more likely to have been reliable. The danger of using an unreliable measure is that it can cause erroneous conclusions. For example, if a test produces wildly different results about the same population of students, in terms of their academic ability, it is unlikely to be reliable. The SAT is said to be a reliable predictor of a student's first year grades in college. This does not mean that the SAT is a fair test, or a valid test of student intelligence or even long-term performance as an undergraduate or in professional school, however, merely that it produces consistent results as a test instrument with different populations over time, subjected by different researchers.

Validity means that conclusions drawn from the data are valid, given the data that was presented, and the way the research was conducted. For example, does the fact that the SAT predicts first year grades in college mean that it tests the 'right' kind of aptitude for college success? This may be a valid conclusion, but there are other possible explanations that question the validity of the instrument, such as the fact that people who do well on the SAT may have demographic characteristics such as their relative wealth that makes it easier for them to have the institutional and family support necessary to succeed during their first year of college. Controlling for other factors, such as working the first year of college, previous quality of education, and so forth, if there is still a correlation between strong performance on the SAT and success the first year of college, this may make the hypothesis that the test is a valid assessment of academic ability to be more accurate. Complications might ensure, however, if certain aspects of the results, such as the fact subgroups such as females do poorer on the test and during the first year of college but do better in graduate school later on, question its validity as a supposedly useful assessment of undergraduate academic potential. It may reliably predict first-year grades but not be a valid predictor of undergraduate success over time.

Unit 3

Question

How does a quasi-experimental method differ from pre-experimental or experimental methods? When might a quasi-experimental design be chosen over an experimental design? What considerations must a researcher make when deciding which approach to take?

Quasi-experimental designs may include a comparison group that does not meet the strict definition of a control, or use separate samplings pre-or post-test to provide different insight upon the research. Even if the findings do not meet strict methodological standards the data may be useful if it is not ethically feasible to create a strict control group and experimental design. For example, say researchers were testing a drug that could save the lives of very sick population. It might not be valid to withhold the drug from the sick population. In a quasi-experiment the researchers could test the individuals taking the drug with and compare the information they collected with information on patients with the disease who contracted the ailment before the drug was developed. This design might be used when comparing AIDS patients before and after retroviral drugs became common. Or a comparison study might be conducted between arthritis patients using a new painkilling drug with those patients using more conventional but acceptable pain-killing treatments like NSAID (non-steroidal anti-inflammatory drugs like Advil).

Pre-experimental designs might involve preliminary research, when the actual research study is uncertain. For example, say the final research will involve subjects involved in a healthy eating and exercise program in an inner city location. To determine what the program components should include and the population that might be most in need of such a program, a case study of the area might be conducted or preliminary statistical data might be gathered about the area. A true experiment consists of a control group and an experimental group, with reproducible results. This is not always the case of a pre-experimental case study or data set, or a quasi-experimental comparison of an actual experiment with data that is not a perfect control for the findings, or was not collected under the same conditions as the research findings. Ethics and a refinement of the research process may require a quasi-experimental or pre-experimental approach.