Any stats experts out there?

torifiletorifile
Posted:
in General Discussion edited January 2014
Ok, I *should* know what I'm doing with statistics by this point in my graduate career, but I'm afraid I've just stumped myself.... Maybe someone could help me.



I've got a number of measures completed by my participants. I've also got multiple assessments on these people. I want to use my measures to predict their number of completions.



Would I use logistic regression? I had used MANOVAs to look at differences between the scores of those who completed once, twice, etc with my number of times as the independent variable, but now I'm afraid that's not right.



Any help? TIA.



edit: my "times" variable is obviously categorical and my measures are all continuous.

Comments

  • Reply 1 of 3
    faust9faust9 Posts: 1,335member
    Took it once and promptly forgot it after the semester was over. Try posting you Q here http://www.physicsforums.com . You'll probably get a quicker response.
  • Reply 2 of 3
    brussellbrussell Posts: 9,812member
    Number of completions doesn't sound categorical, it's a ratio scale, if that's what you want as your dependent measure (0 completions, or 1, or 2, etc.). Sounds like a straight multiple regression with number of completions as the dependent variable and your measures as the predictors. But it may not be appropriate to combine your "completions" in that fashion if they're different things. You could do the separately, each with a logistic regression predicting completion (0 or 1), but then you'd probably want to correct for the fact that you're doing multiple stats by reducing your alpha level or something.



    Also, if you can make your completions continuous (e.g., how far did they get?) that's always going to give you more power.



    Can you be more specific about your measures and assessments?
  • Reply 3 of 3
    torifiletorifile Posts: 4,024member
    Quote:

    Originally posted by BRussell

    Number of completions doesn't sound categorical, it's a ratio scale, if that's what you want as your dependent measure (0 completions, or 1, or 2, etc.). Sounds like a straight multiple regression with number of completions as the dependent variable and your measures as the predictors. But it may not be appropriate to combine your "completions" in that fashion if they're different things. You could do the separately, each with a logistic regression predicting completion (0 or 1), but then you'd probably want to correct for the fact that you're doing multiple stats by reducing your alpha level or something.



    Also, if you can make your completions continuous (e.g., how far did they get?) that's always going to give you more power.



    Can you be more specific about your measures and assessments?



    Thanks, BRussell. I'm going to go a different route because what I really want to know if how people who did varying numbers of sections of my intervention differ from the other groups. That was the reason I did a MANOVA in the first place but I got my head twisted about trying to understand my output... It's not that uncommon for me to get confused when I'm trying to explain my results... I think that for now what I've got will work. Thanks for the advice. I may have to call on you again in about 5 hours.... (no rest for the weary)
Sign In or Register to comment.