1

Basically,

I have been experimenting with STATA for loops and am wondering if I could iterate over every group of four variables to run an alpha test and only print that bundle if alpha is higher than, say, 0.7.

I have 200 different soil and air characteristics and I'd like to know which ones hang together.

thewhitetie
  • 1,057
  • 1
  • 8
  • 7
  • 2
    Are you asking about how to program this? Or is it a question on whether it's a good idea to do this? If the latter, it would help to know what the purpose of doing so is and why you think this might be a good idea. – Björn Jun 17 '16 at 07:12
  • I am asking the former and thanks for the suggestion - I updated my question description. I have 200 variables and I'd like to know which tend to hang together. – thewhitetie Jun 17 '16 at 14:09
  • 1
    What is an "alpha test" and which specific one are you proposing? What exactly do you mean by "hang together"? – whuber Jun 17 '16 at 14:09
  • 'hang together' might be a literal translation from German meaning 'are associated with each other'. I figure 'alpha test' might mean 'computing Cronbach's alpha', which, of course, is not a significance test. – Bernhard Jun 17 '16 at 14:13
  • 2
    @Bernhard Thank you. However, I asked the OP specifically for clarification because I do not want to speculate. If we must speculate about the question, that means different readers will likely understand it differently, leading to misunderstanding of the answers. That is a strong signal that the thread should be put on hold until the question has been edited for clarification. – whuber Jun 17 '16 at 14:29
  • @whuber: I concur – Bernhard Jun 18 '16 at 16:08

1 Answers1

0

If I understand your questions correctly, than factor analysis would be the advisable procedure. The number '4' in your questions seems to be arbitrary and factor analysis will help you find groups of characteristics that are correlated with each other. Principal component analysis might also be considered, if you only want to understand your dataset at hand best. Factor analysis, if you want to generalize conclusions on soils and airs beyond your sample.

Bernhard
  • 8,427
  • 17
  • 38