Correlation and Causation
While causation “Indicates that one event is the result of the occurrence of the other event; i.e. there is a causal relationship between the two. How do we establish a cause-effect (causal) relationship? What criteria do we have to meet? Generally, there are three criteria that you must meet before you. Causality in Social Work Research and Practice However, the association between causality and causal inference in the field of social work and logical.
The relationships described so far are rather simple binary relationships. Sometimes we want to know whether different amounts of the program lead to different amounts of the outcome -- a continuous relationship: It's possible that there is some other variable or factor that is causing the outcome. This is sometimes referred to as the "third variable" or "missing variable" problem and it's at the heart of the issue of internal validity.
What are some of the possible plausible alternative explanations? Just go look at the threats to internal validity see single group threatsmultiple group threats or social threats -- each one describes a type of alternative explanation.Causal Research & Types
In order for you to argue that you have demonstrated internal validity -- that you have shown there's a causal relationship -- you have to "rule out" the plausible alternative explanations. How do you do that? One of the major ways is with your research design.
Let's consider a simple single group threat to internal validity, a history threat. Let's assume you measure your program group before they start the program to establish a baselineyou give them the program, and then you measure their performance afterwards in a posttest.
You see a marked improvement in their performance which you would like to infer is caused by your program. One of the plausible alternative explanations is that you have a history threat -- it's not your program that caused the gain but some other specific historical event.
For instance, it's not your anti-smoking campaign that caused the reduction in smoking but rather the Surgeon General's latest report that happened to be issued between the time you gave your pretest and posttest. How do you rule this out with your research design?
One of the simplest ways would be to incorporate the use of a control group -- a group that is comparable to your program group with the only difference being that they didn't receive the program. But they did experience the Surgeon General's latest report. If you find that they didn't show a reduction in smoking even though they did experience the same Surgeon General report you have effectively "ruled out" the Surgeon General's report as a plausible alternative explanation for why you observed the smoking reduction.
In most applied social research that involves evaluating programs, temporal precedence is not a difficult criterion to meet because you administer the program before you measure effects. And, establishing covariation is relatively simple because you have some control over the program and can set things up so that you have some people who get it and some who don't if X and if not X.
Typically the most difficult criterion to meet is the third -- ruling out alternative explanations for the observed effect. For a conclusive result on causality, we need to do randomized experiments. Why are observational data not conclusive?
We can never conclude individual cause-effect pair. There are multiple reason you might be asked to work on observational data instead of experiment data to establish causality.
First is, the cost involved to do these experiments. For instance, if your hypothesis is giving free I-phone to customers, this activity will have an incremental gain on sales of Mac. Doing this experiment without knowing anything on causality can be an expensive proposal. Second is, not all experiments are allowed ethically.
For instance, if you want to know whether smoking contributes to stress, you need to make normal people smoke, which is ethically not possible. In that case, how do we establish causality using observational data? There has been good amount of research done on this particular issue. The entire objective of these methodologies is to eliminate the effect of any unobserved variable.
In this section, I will introduce you to some of these well known techniques: Panel Model Ordinary regression: This method comes in very handy if the unobserved dimension is invariant along at least one dimension. For instance, if the unobserved dimension is invariant over time, we can try building a panel model which can segregate out the bias coming from unobserved dimension. But, because the unobserved dimension is invariant over time, we can simplify the equation as follows: We can now eliminate the unobserved factor by differencing over time Now, it becomes to find the actual coefficient of causality relationship between college and salary.
And then compare the response of this treatment among look alikes.
Establishing Cause & Effect
This is the most common method implemented currently in the industry. The look alike can be found using nearest neighbor algorithm, k-d tree or any other algorithm. One of them starts smoking and another does not. Now the stress level can be compared over a period of time given no other condition changes among them.
This is actually a topic for a different article in future. This is probably the hardest one which I find to implement. Following are the steps to implement this technique: Find the cause — effect pair. Find an attribute which is related to cause but is independent of the error which we get by regressing cause-effect pair. This variable is known as Instrumental Variable.
Now estimate the cause variables using IV. Try regressing estimated cause — effect to find the actual coefficient of causality. What have we done here?
Statistical Language - Correlation and Causation
Using this methodology, we come out with an unbiased estimation. Now, if we can find any information which is connected to cigarette consumption but not mental stress, we might be able to find the actual relationship. Generally IV are regulatory based variables.
This is amongst one of my favourite choices. It this makes the observational data really close to experimental design. Suppose, we want to test the effect of scholarship in college on the grades by the end of course for students. Because these students are already bright, they might continue being on top in future as well.
Hence, this is a very difficult cause-effect relation to crack! The assumption being that And the only thing which can change is the effect of scholarship.