Hello,
Thanks for all the help. Donald Burrill mentioned that I should
show that r=0.6 is statistically significant vs r=0. I don't think I
have ever seen that done so can anyone please elaborate on how I can
go about doing that. How do I find the test statistic for such a
significance test? Thanks for your help.
As for grouping, I decided to try to separate the cars out by
origin (imported or domestic). Then I can use a significance test to
show if one type is safer than the other for each of the 4 crash test
ratings. I'm also probably gonna carry out that weight grouping
procedure because I am interested in where it will lead me (I will
definately post the results). Here is what I'll do => I have 9 groups
as follows:
Group Number | Weight Range of Vehicle
1 1501 - 2000
2 2001 - 2500
3 2501 - 3000
4 3001 - 3500
5 3501 - 4000
6 4001 - 4500
7 4501 - 5000
8 5001 - 5500
9 5500 - 6000
I take all of my data for 187 cars and split it into the 9 groups.
Then I will find the mean within each group (separately 4 times for
each of the 4 ratings). Lastly I will try to get an LSRL equation that
relates the weight group of the car to its safety rating (not the
specific weight, but the weight group number) for each of the 4
ratings. Maybe this will clear out some confusion from my previous
post. Can such equations be statistically significant? I will do the
above procedure just to see what happens and report back to the group
(I am not necessarilly planning to use that in my report). I think
that I will not get any relationship improvements in front ratings vs.
weight since as some of you have correctly pointed out, there are
numerous reasons for there not to be a specific relationship. However,
do you think I can use this grouping technique for the side ratings
for which I already found that r=0.6 because it seems more logical to
compare cars by weight class rather than by discrete weights? Thanks
so much for all your help.
Cheers,
Serge
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
