Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
So if not statistically different, what are the results showing?
Lets say one group has a salary of 0-20000 and the other has a salary of 20000-40000, are you saying the group that makes twice as much as the other one is not statistically different?
In your example, with two isolated normally distributed groups groups with no overlap between their incomes, the two groups would have a t-value greater than 3, showing a high probability of a statistically significant difference.
If you know this data is almost 20 years old, why did you feel the need to post it? It's 2013. Any report today based on data this old data will be flawed. Its bad when people, in their eagerness to show how "unintelligent" certain groups are, display their own lack of intelligence at the same time.
Ethnicity and income are not independent variables. So you cannot simply say one is more important than the other.
A better approach is to use Bayes rule to draw a network.
The vast majority of GRE takers are white though. So it is always hard to make a convincing analysis.
What are you talking about? Your own source shows that 73% of the GRE takers are white, well that is almost identical to their proportion of the general populace, 72%
The only group truly over represented in the GRE pool are Asians.
As for race and income, if they cannot be separated why did you not mention that in your OP? Because the GRE scores are better predicted by income than by race. And lets not pretend that applying Bayesian modeling supports your idea. It does not. If anything using bayesian method closes the test gap between races because it allows control of certain covariants.
What are you talking about? Your own source shows that 73% of the GRE takers are white, well that is almost identical to their proportion of the general populace, 72%
The only group truly over represented in the GRE pool are Asians.
As for race and income, if they cannot be separated why did you not mention that in your OP? Because the GRE scores are better predicted by income than by race. And lets not pretend that applying Bayesian modeling supports your idea. It does not. If anything using bayesian method closes the test gap between races because it allows control of certain covariants.
Because the GRE scores are better predicted by income than by race.
If they are in the same race.
But if they are not in the same race, no. Please compare low income Asians with upper class blacks, for example.
The reason why income seems to be more important than race is that too many takers are in the same race: white. You understand?
Let me give you a concrete example.
If there are 10 blacks, but 1000 whites. Even if the 10 blacks get extremely low scores, income will still be a better predictor for the whole population, because race can only explain 10:1000, at best. The other 1000 samples cannot be accounted for at all.
But if they are not in the same race, no. Please compare low income Asians with upper class blacks, for example.
The reason why income seems to be more important than race is that too many takers are in the same race: white. You understand?
Let me give you a concrete example.
If there are 10 blacks, but 1000 whites. Even if the 10 blacks get extremely low scores, income will still be a better predictor for the whole population, because race can only explain 10:1000, at best. The other 1000 samples cannot be accounted for at all.
BS. Then RACE would be a better predictor of score. If two variables correlate (SES and race) with another variable (test score) AND each other, it isn't possible for the one with the WEAKER correlation (race to test score) to be the causal factor.
This data may or may not even be relevant anymore. The GRE (and the way it's scored) was substantially changed in August 2011. Looking at data based on the current test might be more helpful.
Ethnicity and income are not independent variables. So you cannot simply say one is more important than the other.
A better approach is to use Bayes rule to draw a network.
The vast majority of GRE takers are white though. So it is always hard to make a convincing analysis.
Do you know how hierarchical regression works? That is the point of it. The IVs with the stronger beta weights are more important than the others, Similarly the IVs that contribute the most change in R2 are more important predictors than those that contribute less to the change in R2.
The best approach would be to run a hierarchical regression both ways.
First:
Step 1: Race
Step 2: Parental Income
Second:
Step 1: Parental Income
Step 2: Race
If the change in R-square is greater in the first regression you can show that parental income explains more of the variance in test scores than race and can therefore conclude it is more "important".
Also there are data transformations to help deal with skewed data. In the Quant. example the reflection of the natural logarithm would be a good start. The verbal is gaussian so there is no need for data transformations.
It would also be interesting to see if the overall score is gaussian.
Quote:
Originally Posted by Vergofa
To provide information and initiate discussion.
Again, the discussion should be around the overlap of test scores that are within one standard deviation of one another. The standard deviations within groups are huge. Meaning people in all races score all over the place.
For example, I am a white male and I scored a 1300; 560 Verbal 740 Quant. Had practice scores of 1360: 600 Verbal 760 Quant. So what does that mean say about me?
The discussion would be more meaningful if the standard deviations were smaller, but they are not.
Please register to post and access all features of our very popular forum. It is free and quick. Over $68,000 in prizes has already been given out to active posters on our forum. Additional giveaways are planned.
Detailed information about all U.S. cities, counties, and zip codes on our site: City-data.com.
Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
The point
