Unable to determine the right test?

iha2005

New Member
hello,
I'm doing a paper on measuring authors productivity and unable to decide what should I use as my statistical test to come up with solid results. Here is what the initial results look like (this is an example, I have more the 2000 authors):

Rank Productivity Rank Productivity Authors
4 10 papers 1 40 papers A
1 40 papers 2 35 papers B
2 25 papers 3 30 papers C
3 20 papers 4 25 papers D
5 5 papers 5 20 papers E

it is ordered by author names, number of papers authored then the rank, and then with another way of calculating productivity for the same authors and the rank on that way.
please note that after about 500 authors each will have only one paper and the same rank in both ways.
what do you think is the best way to compare, rank correlation, chi square, or what... please advise...thanks for the help

hlsmith

Not a robit
What is your study question and hypothesis? Are you attempting to compare the two ranking methods?

iha2005

New Member
yes I'm, I have two measures H and M and want to know if there is statistical difference between them, each one is calculated in a certain way and gives a rank and the ranks are close that's why I want to test them statistically.

CB

Super Moderator
yes I'm, I have two measures H and M and want to know if there is statistical difference between them
I think you still need to be clearer on what you are trying to find out. What, specifically, do you mean by a "statistical difference"?

Obviously the methods do not always provide the same ranks. If that's all you wanted to find out, you are finished.

Or are you actually interested in how closely related the rankings of the two methods are?

The key to good analysis is knowing what you are trying to find out.

noetsi

Fortran must die
The key to good analysis is knowing what you are trying to find out.
That should be written in gold at every university in the land.

iha2005

New Member
I think you still need to be clearer on what you are trying to find out. What, specifically, do you mean by a "statistical difference"?.
One method is very simple and the other is very complicated, what I want to know is do I really have to use the complicated one, if there is no difference in the rank why do I have to torture myself? I already have the two rankings but I don't think I'm done. so I want to know I they are so close there is no need to get complicated, if not the second method would be a must for me. Does this mean measuring how close the rankings are? thanks

CB

Super Moderator
if there is no difference in the rank why do I have to torture myself? I already have the two rankings but I don't think I'm done. so I want to know I they are so close there is no need to get complicated
Ok, well as I said we can already rule out the hypothesis that there is no difference between the rankings. There are some differences just in the sample data.

So the question is: How closely related are the rankings of the two systems? And how closely related do you need them to be?

You can use a correlation method for ordinal data such as Spearman's rho to see the relationship between the rankings. The thing you will need to decide is: How high do you need this correlation to be for you to decide method 2 is acceptable? That is something for you to decide.

Hope that helps!

iha2005

New Member
Thanks for the help Cowboybear, I'll Use the Spearman test and let you know how it goes