User:Aaron1509/Causal Explanation and Research Design

The major challenge of social science is to arrive at scientifically valid explanations for why we see the particular patterns in the social world that we actually observe. This effort is fraught with challenges of measurement, analysis and inference. This article is about how studies can be designed to provide the best evidence for revealing the causes of social outcomes and patterns.

Research and research design are built into a causal explanation which results in an explanation of the social world.

Nomothetic Causal Explanation
A Nomothetic causal explanation exists when there is a correlation between an Independent variable and a dependent variable. That is to say the value of the dependent variable would be different from what it would be if the independent variable occurs or acts upon it. Social Scientists use the nomothetic method when they have an interest that involves social regularities or things that apply to people in general. The data involved in Nomothetic causal explanation is quantitative data, derived from numbers.

A Social Scientist applying the nomothetic-deductive method may run into several problems. For instance this type of research cannot rely on information on any one individual; it needs information that describes general trends, patterns, or relationships of many cases. Another problem is that this type of research would work best if information on all people, events, and places, could be known. Since this is impossible a Social Scientist is forced to rely on Inferential Statistics, logical tools mixed with information about a population which is then determined from information on a portion of the given population.

Idiographic Causal Explanation
An idiographic causal explanation is a scientific explanation that includes a sequence of events that lead to a particular outcome for a specific individual. Stated idiographically, a causal explanation would include initial conditions and then would relate a series of events at different times that led to the outcome. Social Scientists using the Idiographic method are concerned with how a specific result occurrs as part of a larger whole or larger set of circumstances that are related. This type of causal explanation is concerned with an understanding of human behavior. This explanation is also more concerned with individual people, places, and events rather than the general population. This concern is also somewhat of a problem. The social researcher can make general conclusions but only based on an indivual, a single place, or a single event. This means that the idiographic method can not be used to explain any general ideas, places, events, or populations. The information gained is limited to the context of just that specific person, place, event.

Criteria for Causal Explanations
When designing research techniques, social scientists use five criteria to help decide the accuracy of the results. When research is finished, the researchers need to go back over the data and use the five criteria to help better understand the results. If one of the criteria is not met while conducting research, we find it difficult to believe the validity of the research. The five criteria are correlation, time order, nonspuriousness, causal mechanism, and context. Correlation, time order, and nonspuriousness are the most important of the criteria, however, causal mechanism and context can also stregthen causal explanations.

Correlation
Correlation basically establishes that if variation in one variable occurs, then variation in the second variable should follow suit. If values in the independent variable differ in the same terms as the dependent variable, then correlation exists. This test is the same for experimental and non-experimental research. The only difference is that in non-experimental research, the independent variable is not the treatment. To establish a correlation the measurements of the data must be valid, meaning measuring what is supposed to be measured. Validity refers to finding results that correctly show the concept being measured. If a social researcher can ensure a valid measure then the data retrieved is reliable and a correlation is established. If the data found is covering more information than what is necessary it may not show a correlation.

Time Order
Time order is also an important criterion for causal explanation. Time order tells us that the variation in the dependent variable did, in fact, occur after variation in the independent variable. Basically, time order says whatever causes the outcome actually has to happen before the outcome. A causes B, but we have to make sure that A actually happened before B. Establishing time order is essential because this will determine the research design and help create causal explanations. Two basic researchs designs are longitudinal and cross-sectional. The problem that occurs is that a longitudinal research design is the only proper way to establish time order. Longitudinal design occurs at more than one point in time which allows the researcher to ensure that A did indeed occur before B. Due to the fact that cross-sectional design involves gaining data at one point in time it becomes impossible to establish which occured first, A or B, and therefore making it impossible to establish time order.

Nonspuriousness
The concept of nonspuriousness draws from the idea of time order. While time order tells us that A absolutely caused B, nonspuriousness tells us something a bit different. In some cases it may be true that A caused B, but maybe a third variable was involved, C, that caused both A and B. In this case A would not have caused B at all. The phrase "correlation does not prove causation" helps us with this concept. This criterion for causation is especially important. It would be easy for an entire set of research to become invalid simply because a third variable was overlooked. A way to deal with nonspuriousness is to consider sampling frame. Sampling frame sets up the boundries for what will be included in the research. When conducting research, we are able to consider what should be included in our study if it fits into the boundries that we have set. By setting the boundries of the sampling frame, a researcher is able to control which causes are studied. Therefore, the researcher can easily check for a nonspurious relationship.

Causal Mechanism
A Causal Mechanism is a connection between the variations in the independent and dependent variables that is created by social mechanisms. In a nutshell, a causal mechanism is what causes the relationship between the independent and dependent variables. If there is no connection between the the variables, then there is no causal mechanism. Intervening variables help explain the connection between the independent and dependent variables. The intervening variables truly explain the variance between the independent and dependent variables. The only drawback in the intervening variable is that finding it does not guarantee any help in finding the causal mechanism. For any experiment, there can be numerous amounts of causal mechanisms. There is no set number of causal mechanisms in an experiment.

Context
Context from a sociological view is basically what the data is about. The context of the data in an experiment needs to deal directly with what the experiment is supposed to be finding. Stemming from context is the contextual effect, which is the reason why something happens differently in different sociological and geographical settings. Identifying the contextual effect can clarify the relationships between the independent and dependent variables. Different types of contextual effect are: race, age, financial status, geography, gender, The list continues on to many other types of contextual effects.

Cross-Sectional Design versus Longitudinal Design
Cross-sectional design can be defined as a measure of the actions, attitudes and characteristics of respondents at one point in time. A cross-sectional study takes place at a single point in time and observing the events that are occurring at that moment. Some attributes of cross-sectional design are cost and time efficiency. This type of study can be done quickly and observe a large number of participants with few cost constraints. Furthermore, a researcher does not need to worry about participants dropping out of the study, since it only covers one particular point in time. With information from only one point in time, it is difficult to explain cause and effect; correlation is easy to identify with cross-sectional design. The problem here is that the particualar information is only gained over one point and time and aspects of individuals, and events may change.

The social researcher may use longitudinal design in an effort to note changes over time. Longitudinal design takes place over a period of time, not at one particular moment as in cross-sectional design. This design takes measurements at least two different times. The problem encountered here is that it becomes difficult to make a serious generalization about a population with only one sample.

To deal with this problem a social scientist would do another type of logitudinal design called repeated cross-sectional design (trend study, repeated measures, time study). This allows the social researcher to have many samples over time in order to recieve a better generalization of a population but this comes at the cost of the amount of information gained in a standard longitudinal design. A repeated measures test is when measurements are taken at least two times, but only taken a few times. It turns into a trend study or time study when the test is performed at least twenty times.

Resources

 * King, Gary, Robert O. Keohane, and Sidney Verba. Designng Social Inquiry. 1st ed. Princeton: Princeton University Press, 1994.


 * Kelley, Debra. "Causal Reasoning." Causal Reasoning. 13 Oct 1999. 3 	Oct 2007 	.


 * Schutt, Russell K.. Investigating the Social World. 5th. London: Sage 	Publications,      2006.


 * Singleton, Royce, A., and Bruce C. Straights. Approaches to Social Research. 3rd ed. New 	York: Oxford University Press, 1999.


 * Trochim, William K.. "Time in Research." Social Research Methods. 20OCT2006. 25 Oct 2007      http://www.socialresearchmethods.net/kb/timedim.php.
 * http://www.sfu.ca/~richards/Zen/Pages/Chap1.htm