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.
Again the standard deviations are the more important data point.
Look at it in terms of effect sizes.
Most researchers suggest an effect size of .20 is a small effect size, .50 is a moderate effect size and .80 is a large effect sizes.
Let's look at white and asian quant for example.
569 vs. 622. with a standard deviation of 131 means that the effect size is just over .50, so it is a moderate effect size, or the numbers are moderately different.
Just because means are different doesn't necessarily mean the numbers are drastically different. Again, the standard deviations (which are huge) are the more important data point.
The only really huge differences are between AA and White and Asian in Quant. Other than that all are basically within 1 SD of each other.
If you were doing a T-test to see if these groups were different from each other almost all of them would turn out to have a T-value of less than 1 and all would have T-values of less than 2.
Standard statistics would say you need a T-value of at least 1.65 to consider the groups statistically different. It looks like none of these differences would meet that threshold.
If you were doing a T-test to see if these groups were different from each other almost all of them would turn out to have a T-value of less than 1 and all would have T-values of less than 2.
Standard statistics would say you need a T-value of at least 1.65 to consider the groups statistically different. It looks like none of these differences would meet that threshold.
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?
Still wondering what the point of the OP was. Did you have a comment? a question?
Mizzourah2006, you make excellent points about statistical significance.
I also would like to see what the scores are after controlling for social class (maybe the longer report does that, but all I looked at were the numbers). I would suspect that much (although likely not all) of the racial difference goes away. Anyone read the research on "stereotype threat"?
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?
He's talking about STATISTICAL SIGNIFICANCE. Might I suggest a stats class?
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?
If one group's mean is salary is 20k but their standard deviation is 10k and another group's salary is 30k with a standard deviation of 10k, I am saying that the groups themselves aren't statistically different, yes. That's why SDs are so important.
A mean of 20k with an SD of 2k tells us a lot more about the group than a mean of 20k and an SD of 10k. The 10k SD means that there are probably a lot of people that make a good salary and a lot that make almost nothing. The SD of 2k tells us that they are all right around the the mean of 20k.
So in your example if the mean is 20k and the SD is 20k for group 1. and the mean is 40k and the SD is 20k for group 2 they are not that different, yes. Again if group 1 is 20k and the SD is 5k and group 2 is 40k and the SD is 5k they are different.
If one group's mean is salary is 20k but their standard deviation is 10k and another group's salary is 30k with a standard deviation of 10k, I am saying that the groups themselves aren't statistically different, yes. That's why SDs are so important.
A mean of 20k with an SD of 2k tells us a lot more about the group than a mean of 20k and an SD of 10k. The 10k SD means that there are probably a lot of people that make a good salary and a lot that make almost nothing. The SD of 2k tells us that they are all right around the the mean of 20k.
So in your example if the mean is 20k and the SD is 20k for group 1. and the mean is 40k and the SD is 20k for group 2 they are not that different, yes. Again if group 1 is 20k and the SD is 5k and group 2 is 40k and the SD is 5k they are different.
but practically, they are hugely different, because almost everyone in group 2 will be making twice as much as almost everyone in group 1.
let me give a clearer example of my point. Lets say group a is exposed to radiation levels at mean 10 with a standard deviation of 5 and group b is exposed to radiation levels at mean 15 with a standard deviation of 5. lets say the threshhold of deadly radiation is 15. half of group b is going to be dead, just about everyone in group a is going to be ok. There is clearly a statistical significance in predicting who is going to die. You cant just say '1.6 is the t-value we want in order for it to be significant' that is pretty meaningless without the context of what the range of values represent and what you want to know with the statistics
but practically, they are hugely different, because almost everyone in group 2 will be making twice as much as almost everyone in group 1.
Build out the distribution:
Group 1:
68% earn between 0-40k.
95% earn between 0-60k
99% earn between 0-80k
Group 2:
68% earn between 20-60k
95% earn between 0-80k
99% earn between 0-100k
Essentially both groups are all over the place. The difference is it looks like there are probably a lot of people that earn 0 dollars in group 1 and fewer that earn 0 dollars in group 2.
Quote:
Originally Posted by HiFi
let me give a clearer example of my point. Lets say group a is exposed to radiation levels at mean 10 with a standard deviation of 5 and group b is exposed to radiation levels at mean 15 with a standard deviation of 5. lets say the threshhold of deadly radiation is 15. half of group b is going to be dead, just about everyone in group a is going to be ok. There is clearly a statistical significance in predicting who is going to die. You cant just say '1.6 is the t-value we want in order for it to be significant' that is pretty meaningless without the context of what the range of values represent and what you want to know with the statistics
that's the difference between practical significance and statistical significance. Those groups really arent' that different in their exposure to the IV, the difference is there is a threshold with which that exposure becomes deadly.
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.
