Reply With QuoteDid you ever figure this out?
Hi. I was excited to find this forum and appreciate your trying to help me with my stats question. Here goes: I am a sports psychologist studying various aspects of football performance to try to determine what are the most influential variables in explaining which team wins the game and/or performs better.
I have collected data from 50 important but random NFL games, and this data consists of values for 37 different variables. So my correlation matrix, which I have placed on SPSS, consists of a huge table showing the Pearson Product Moment correlations of 37 variables and how they correlate with the other 37 variables. To give you an example of some of the variables, so you will know what I am talking about, they are variables such as:
1. total yards gained
2. number of penalties
3. time of possession
4. total number of turnovers
So there are 50 games which equates to 100 teams in the data matrix and each of these 100 teams has a particular value for each of these 37 variables in this huge matrix. So I have the means and standard deviations of each of the variables, and that is interesting in that it allows me to describe the data well and create z scores and percentiles to show how individual teams did compared with the other 99 teams, and this is interesting, but what I would like help from you on is a method to go about determining what wins football games. It so happens that of the 37 variables, 2 are perfect for outcome, or how a team did. The first is simply win or lose as a categorical variable with teams assigned a 1 for a win and a 0 for a loss. I have been told that this is appropriate to look at the correlations and that it can be treated as if it is a pearson product moment statistic or "r" ... and the other outcome variable is the point differential between the winning and losing team (so a number between 1 and 50 for example). So I want to know, using SPSS, what kinds of analyses to run to be able to know which of the 36 variables best correlate with outcome. I can look at the correlations and that is very interesting, but as we all know, that is not enough. I need to know if the correlations are statistically significant (meaningfully differerent from zero) and also if the correlations are different between variables too ... but mostly I need to do some kind of overall analysis that tells me which of the variables most correlate with outcome. This might be an anova or a factor analsis ... but I am not sure and just need some conceptual clarity ... and using SPSS. So in sum, this is a very simple question of a huge correlation matrix with 36 numerical variables and how they stack up against one another in correlating with outcome. For example, how would I find out the top 3 or 4 correlates of outcome, and how confidently could I also say that one is more important than another. I hope my question here is clear. It seems very straightforward, and I just need some methodological guidance. Thank you very much!
Did you ever figure this out?
Posting Permissions
|
|