Notions of Correlation and Causation - Essay Example

At present, we see, hear, and read a lot of stories reporting new discoveries practically on the different aspects of our lives like health, well-being, relationships, etc. Moreover, most of those stories (especially the articles that we read on magazines) use words denoting causation (e.g., smoking “causes” this and that), which may really spark the interest of readers like us but there is a line separating what we call causation from correlation and that they are really not the same. “The Concise Oxford Dictionary defines correlation as mutual relation between two or more things and causation as causing or producing an effect” (“Correlation and Causation,” 2001). If two variables are correlated, it still does not mean that one variable causes the other to vary as it does even if the statement makes sense (Jaccard & Becker, 2002; para. 1, “What is the difference,” n.d.). If one action causes another, then they are most certainly correlated therefore causation causes correlation and not the other way around (Deutsch, 2005; para. 1, “What is the difference,” n.d.). Moreover, in using correlational data, causal inferences cannot be made even if we obtain a perfect correlation which may be a +1.00 or -1.00 (Myers & Hansen, 2006). If causal inferences are to be drawn from correlational analyses, extreme caution must be made (Jaccard & Becker, 2002). Actually, there are four possible reasons as to why two variables X and Y might be correlated. Four possibilities are that (1) X causes Y, (2) Y causes X, (3) X and Y affect each other which is known as bidirectional causation, or (4) some additional variable(s) causes both X and Y (Jaccard & Becker, 2002; Myers & Hansen, 2006). To further illustrate these possibilities, let us explore some examples. For illustration purposes, let us say we find a positive correlation between the number of hours college students spend working for pay and the number of campus organizations college students belong to, it is unlikely that working causes students to join organizations or that membership in organizations causes students to work but the correlation between hours of work and group membership is probably attributable to students’ desire to achieve and related personality characteristics (Jaccard & Becker, 2002). There are also examples wherein the causal relationship underlying a correlation is ambiguous such as the correlation between the amount of violent television a child watches and child’s aggressiveness. In this case, there are four possible causal directions: (1) watching violent programs on television causes a child to be more aggressive (2) higher levels of aggressiveness cause a child to watch more violent programs on television (3)higher levels of aggressiveness cause a child to watch more violent television, and, at the same time, watching more violent television causes a child to have higher levels of aggressiveness (4) an unmeasured third variable such as low autonomic arousal causes a child to watch more violent television programs and causes a child to have higher levels of aggressiveness (Myers & Hansen, 2006). It has already been mentioned that when two events correlate, it does not follow that one has caused the other. The Latin term for an error that we commit when we are confused between correlation and causation is called non causa pro causa which means non-cause for the cause (para. 2, “Faulty Causation”, n.d.). Our intuition may be blamed for committing this error because it can lead some of us astray when it comes to distinguishing between causality and correlation (para. 2, “What is the difference,” n.d.). However, more than our intuition, the words chosen by media can also be blamed for the confusion. “The media often concludes a causal relationship among correlated observances when causality was not even considered by the study itself” (para. 11, “What is the difference,” n.d.). As much as possible, we must always use our critical thinking in evaluating different articles or stories that we encounter in our daily lives. However, according to Gardner (1999), under certain conditions, correlation can imply causation. He said that “correlation does imply causation if the independent variable is randomly assigned to the participants” (p. 14). Random assignment is very important in experiments therefore there may be times wherein correlation and causation may be difficult to separate. There are various research methods that we can choose from when doing a research study but only a true experiment (most likely a laboratory experiment) permits us to make causal statements (Myers & Hansen, 2006). The type of cause and effect relationship established through experiments is called a temporal relationship but the fact that one event precedes another is not alone sufficient, even the Scottish philosopher David Hume said that we can never establish causality from temporal relationships (Myers & Hansen, 2006). Other kinds of relationships can suggest cause and effect namely spatial and logical but they do not appeal that much to researchers because other potential causal explanations are never completely ruled out (Myers & Hansen, 2006). Internal validity is “the degree to which a research design allows us to make causal statements” (Myers & Hansen, 2006, p. 62). A causal relationship between variables may be determined using carefully designed experimental procedures as mentioned previously, but since there are problems which cannot be answered experimentally because of practical and ethical concerns (e.g., pregnancy, smoking, alcoholism, or homelessness) nonexperimental designs such as a correlational design might be utilized instead of an experimental design. Correlational designs can show relationships between sets of antecedent conditions [circumstances that come before the event or behavior that we want to explain] and behavioral effects, but the antecedents are neither manipulated nor controlled by the researcher (Myers & Hansen, 2006). According to Myers and Hansen (2006), in a correlational study, selected traits or behaviors of interest are measured first wherein the numbers (i.e., scores) are recorded which represent the measured variables. Following this, the degree of relationship or correlation between the numbers is determined through statistical procedures wherein the Pearson Product Moment Correlation Coefficient (r) is the most commonly used statistical procedure for calculating simple correlations. The values of a correlation coefficient can vary between -1.00 and +1.00; the sign tells us the direction of the relationship while the absolute value of r tells us something about the strength of the relationship (Myers & Hansen, 2006). Furthermore, once the correlation is known, we can already make predictions that is if we know a person’s score on one measure, we can make a prediction of that person’s score on another measure that is correlated to it and that the higher the correlation, the better our prediction will be (Myers & Hansen, 2006). Other correlation-based research designs are multiple correlation which “determines the degree of relationship between three or more variables,” linear regression analysis which “uses regression equation to predict scores on one variable from scores on a second correlated variable,” multiple regression analysis which “uses regression equation to predict scores on one variable from scores on sets of other variables,” and factor analysis which “determines subsets of correlated variables within a larger set of variables” (Myers & Hansen, 2006, p. 142). As mentioned previously, one drawback of correlational designs is the direction of cause and effect but there are already tools for causal modeling in correlation-based designs namely path analysis and cross-lagged panel designs (Myers & Hansen, 2006). Path analysis “uses beta weights from multiple regression analysis to generate possible direction of cause and effect from correlated variables” while cross-lagged panel “measures the same pair of variables at two different points in time and looks at patterns of correlations across time for possible direction of cause and effect” (Myers & Hansen, 2006, p. 142). Even if correlational designs are low in internal validity, they tend to be higher in external validity [applicability to people and situations outside the research setting], than laboratory experiments (Myers & Hansen, 2006). References Correlation and causation. (2001). Retrieved November 9, 2008, from http://www.cambridge2000.com/memos/correlation.html Deutsch, J. (2005). Correlation vs. causation. Retrieved November 9, 2008, from http://www.statistics-help-online.com/node50.html Faulty causation fallacies: Correlation vs. causation (non causa pro causa). (n.d.). Retrieved November 9, 2008, from http://atheism.about.com/library/FAQs/skepticism/blfaq_fall_correlation.htm Gardner, R.C. (1999). Correlation, causation, motivation, and second language acquisition. Canadian psychology, 41, 10-24. Jaccard, J., & Becker, M. A. (2002). Statistics for the behavioral sciences (4th ed.). Belmont, CA: Wadsworth. Myers, A., & Hansen, C. (2006). Experimental psychology (6th ed. ). Belmont, CA: Wadsworth. What is the difference between causation and correlation? (n.d.). Retrieved November 9, 2008, from http://stats.org/in_depth/faq/causation_correlation.htm Read More