+ Reply to Thread
Results 1 to 2 of 2

Thread: Variable prediction

  1. #1
    Points: 39, Level: 1
    Level completed: 78%, Points required for next Level: 11

    Posts
    3
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Variable prediction




    Hi everyone,

    I'm currently doing work with morphometry, working with data I've collected and with an already existing database. At the moment I'm stuck at trying to make both databases comparable, and that's the reason I'm asking for your insight.

    My problem is this: I have a set of 9 continuous variables (lets call them A, B, C...I), all of them related, that were measured in two different ways in my data and the other database - I only received it and noted the differences after data collection. This means that I have database 1 with measurements of variables A-I, and database 2 with measurements of variables A*-I*. Although the measurements were of the same morphological traits, the different measuring techniques make it rather difficult to compare.

    I'm still collecting data for my database 1, and now measuring the traits both ways so I have a pool of data for comparison, but a problem still remains, because the two measuring techniques tend to give slightly different proportions to each variables - it is not a direct correlation in some cases.

    To summarize, I'm looking for a way to transform A->A*, B->B*, C->C*... etc. given the constraints of each measuring technique, using (or not, if possible) the data pool where both measurements were taken. Can anyone help?

  2. #2
    Points: 39, Level: 1
    Level completed: 78%, Points required for next Level: 11

    Posts
    3
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Variable prediction


    Hey guys, I'm going to give it one more try if you don't mind!

    I've continued to work on this problem, and I' ve tried different solutions, but I'm still looking for something better than what I have. I've already tested simple linear regressions and multiple linear regressions, and the correlation is seldom acceptable.

    I've encountered recently some functions on my analysis software (StatSoft's Statistica) that seem promising, namely principal component regression models. Can anyone help me in regards to understanding a bit how I could use them for my problem?

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats