Elaboration and Rosenberg Summary Part Two

The thoeretical purpose of introducing an antecedent variable is the same as that of introducing an intervening variable, namely, to trace out a causal sequence.

ANTECEDENT VARIABLES is one which comes BEFORE the Independent Variable in the sequence.

Note that the ANTECEDENT VARIABLE is a true effective influence; it does not explain away the relationship between the Independent and Dependent Variables, but clarifies the influences which preceed the relationship.

So say you have discovered a causal relationship between Education (Independent Variable) and Political Knowledge (Dependent Variable). You have found evidence of causality and everyone agrees with you and all is well in the world. But, then you get bored and your 15 minutes of fame are nearing their end so you decide to get your ass back in gear and build upon your already solid relationship. What ever will you do? How about trying to find out what causes your Independent Variable (Education)? This is the question that ANTECEDENT VARIABLE analysis is designed to answer.

Now... yippee! You've found support for a causal relationship between Social Class (Independent Variable) and Education (Dependent Variable). In the relationship between Social Class and Education and Political Knowledge, Social Class is the ANTECEDENT VARIABLE. You are in the spotlight once more and are the talk of the town (for a little bit longer...)

In order to fully qualify as an ANTECEDENT VARIABLE, there are three statistical requirements which must be satisifed:
1. All three variables - antecedent, independent, and dependent - must be related. (That only makes sense! And is just a little bit obvious.)
2. When the ANTECEDENT VARIABLE is controlled, the relationship between the Independent and the Dependent Variable should not vanish. (Well, no shit! This relationship had already been established before we introduced the antecedent vairiable, it sure as hell better not disappear when we put things back the way they were!)
3. When the Independent Variable is controlled, the relationship between the Antecedent Variable and and the Dependent Variable should disappear. (This makes sense if you remember how that worked with the Intervening Variable. In this case, the Independent Variable is in the middle - like the Intervening Variable was - and just like before, it would be like taking the meat out from in between our slices of bread - no more sandwich!)


Rosenberg Chapter 4

SUPPRESSOR VARIABLES are nasty little buggers that can fool you if you don't look out for them and uncover their existence. They sneak into your study/research, doing their best to remain undetected, and work to "cover up" and/or "hide" things that directly affect the apparent relationship between your Independent and Dependent Variables.

Basically, you need to know that it is possible to be misled into thinking that there is seemingly no realtionship between two variables when there really is one. It may appear that there is not an inherent link between the variables, whereas, in fact, the fact that you cannot "see" the relationship (or that there there is an absence of any relationship) between the Independent and Dependent Variables may be due to the intrusion of a third variable.

Remember the overhead example with the study that wanted to measure Feelings of Estrangement in African-Americans as compared to Caucasion peoples? The simple analysis done in a Bivariate Table showed "no" difference between culture and estrangement, but then they controlled for education and then we could see that whites had a higher percent of cultural estrangement than african-americans when basic v. higher education was controlled for.

All in all, SUPPRESSOR VARIABLES are like the "opposite" of EXTRANEOUS VARIABLES. Remember, Extraneous Variables (storks and babies) made it look like two variables were related, but they weren't and when the Extraneous Variable was controlled for the relationship disappeared. Well, in this case, when the SUPPRESSOR VARIABLE is controlled for (tah-dah!) the relationship is revealed.


Now, if you thought the Suppressor Variables were menacing, wait until you hear about these DISTORTER VARIABLES. When you find one of these (hidden deep within your analysis, undetectible in Bivariate Tables just like the Suppressor Varaibles were) they reveal that the correct interpretation is (now watch this) precisely the reverse of that suggested by the original data. These rat bastards convert a positive relationship into a negative relationship.

Rosenberg (and I think Nigem, too) gave the example of a study done in which it was found that lower-class people are somewhat more likely to hold favorable attitudes towards civil rights. Researchers thought that could mean that lower-class people have a generally more "liberal" or "progressive" idoelogy (favoring civil rights). They then went on to speculate on the bearing of an underprivelaged social position on an ideology favoring equal rights. (Now for the big switcharoo...) The study was done in Washington D.C. (a city with a high African-American population) and when they examined the relationship of class to civil rights among Whites and African-Americans seperately they found (guess what?): the relationship is exactly opposite of that originally shown. In reality, among African-Americans, upper-class people are more likely than lower-class people to favor civil rights and the same is true for Whites.

There you have it - Rosenberg, Cliff's Notes Style

It's 3:08 am... do you know where your Methods Project is?