Test of Significance?

#1
I want to conduct a significance test in finding the significance between either a C+ vs a B- in AP statistics class will affect chances of attending my target university only in accordance to GPA.

I will use the average GPA of that school to determine this.

How can I go about with this procedure? Does it make sense to you? What test should I use? Do u think their will be significance at the .01 or .05 alpha levels?
 
#4
Ill give an example:
say a certain university's average accepted gpa is a 3.5. I will assume that obtaining a 3.5 gpa will get me into that university. I want to find if their is a significant difference if in one class I get a C+ instead of a B-. I have 7 classes so I will average out my total gpa from each class. with a C+ in AP stat my gpa is 3.47 and with a B- my gpa is 3.52. So I want to see if my chances are significantly greater if I have a B- instead of a C+.
 

Dason

Ambassador to the humans
#5
Well it depends on the university and their criteria. It depends if the class was a course in your major. Do you have any replication?

I mean just by common sense you should know that a better grade will increase your chances but it doesn't guarantee you anything. For instance there are a lot of 4.0 students that don't get into MIT...
 

bryangoodrich

Probably A Mammal
#6
I don't see that there is really a stochastic relationship here. If you assume a 3.5 will get you into the university, and a higher grade gives you a higher average GPA, then it is a deterministic relationship by your very assumption (and the definition of a mean) that a B- raises your chances. When you ask about a "significantly greater" chance, I don't see where chance comes into the question. Either B- improves your average or it doesn't, and a B- never gives you a lesser average than a C+ would, they are ordinal (ordered) values: the B- is always of greater value than a C+. In other words, a B- is monotonically increases your mean GPA compared to a C+. How, then, would we define significance in such a way to say whether a B- does a better job than a C+? It apparently always does a better job, but significance in the way you are referring does not seem to be statistical significance. If you think otherwise, could you elaborate?
 
#7
Is their anything statistical that I can prove using like a t-test or something? I want to find how much more likely it is to get into a certain university based solely on gpa. Obviously a higher gpa will give you better chances, but I want to see how much better the chances are, how can I do this?
 

Dason

Ambassador to the humans
#8
It would take quite a bit of work to flesh something like that out completely. Seriously for your interests your probably just better off asking somebody that works in admissions. Also just note that while having a better grade will improve your chances slightly there are things on the application that matter so much more than GPA that the difference between a C+ and a B- in AP Stats isn't going to matter too much.

It would take quite a bit of work (and a long time) to answer your question in a sound statistical manner in such a way that you could show that that slight increase in GPA was the cause for an increased probability of admission.
 

Link

Ninja say what!?!
#9
I think there's a statistical model that you can put this question into. HOWEVER, there are a lot of covariates that you'll have to take into consideration. Once you take all the other variables into account, I can see a more stochastic relationship. Some examples are state residency, major, community service, SATs, etc.
 

Dason

Ambassador to the humans
#11
Hehehe. Dason beat me to it. But yes, it would require quite some work.
And then I edited my post thinking nobody would reply in that amount of time making you look like a ninja quoting out of nowhere!

... The reason I edited it was that I reread the first post and it just seems like the OP is a high school student who is worried about getting admitted to their target university and is hoping for statistics to give them a magical answer.
 
#12
I think there's a statistical model that you can put this question into. HOWEVER, there are a lot of covariates that you'll have to take into consideration. Once you take all the other variables into account, I can see a more stochastic relationship. Some examples are state residency, major, community service, SATs, etc.
No I want to disregard all confounding variables, I only want this to be about gpa. So a 3.47 against a 3.52, and the average gpa of the university is 3.5 I want to see how much better the chances are if I would have a 3.52 instead of a 3.47 disregarding all confounding variables (SAT, ACT, community service hours, majors, etc...)

And I want this to be done using statistical procedures, im hoping their is such a test i can use Ex: T-test, Chi squared, etc.... I dont know which test to use...
 

Link

Ninja say what!?!
#14
Disregarding all confounding variables means your answer WILL mislead.

Take an example. Lets say I want to see if matches cause lung cancer. Reason and logic tell us probably not. But lets assume we don't know that. So we gather a sample and do a comparison NOT controlling for confounders. We would see that matches ARE associated with lung cancer.

The reason is because matches are highly associated with smoking. Smoking is highly associated with lung cancer. Therefore, if you don't control for smoking, you'll get a misleading answer.