+ Reply to Thread
Results 1 to 5 of 5

Thread: Determining the weight of variables

  1. #1
    Points: 8, Level: 1
    Level completed: 15%, Points required for next Level: 42

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Determining the weight of variables




    I am trying to determine if one of the numbers in regression summary provide the weight of the important of each variable?

    Here is an example:

    Say a sales manager wants to determine how much of a reps time should be allocated on the variables that predict whether a sale is made. Let's say the dependent variable is whether a sales is made, and the independent variables are Sales Training, Number of Prospecting Calls Made, and Sales Presentation. Is there any way to look at the regression summary to get an idea of the impact of each variables so you could see percentages such as Sales Training accounts for 50% of the prediction, Number of Prospect Calls 40% and Sales Presentation 10%.

    Would the t value provide that? If I divided the t value of an individual variable by the sum of the t values for all the variables would it give me a percentage that would roughly tell me the importance of variable? I'm hoping you can point me in the right direction.

    Thanks,

    Mike

  2. #2
    TS Contributor
    Points: 6,786, Level: 54
    Level completed: 18%, Points required for next Level: 164

    Location
    Sweden
    Posts
    524
    Thanks
    44
    Thanked 112 Times in 100 Posts

    Re: Determining the weight of variables

    The partial coefficient of variation of partial correlation coefficient may be what you are looking for. It doesn't completely answer your question though. In almost all fields the coefficient of variation is misused. It gives an estimated percentage of variation that is explained by one or more variables, and thus it is not a coefficient of determination, so to speak.

    An illustrative example of this problem is to consider body length. Variation in length is probably explained by genes to a degree of, say, 0.9. And the rest of the variation is explained by non biological factors. But this doesn't per se mean that length is determined by a degree of 0.9 by genes. It is still possible to alter your length extensively by altering non biological factors. For example, cut off your legs or stand below an elevator that smashes into you.

    What do I want to say with this? Well, don't misinterpret a coefficient of variation by saying that "Y is determined by X to a degree of 90 percent".

  3. #3
    TS Contributor
    Points: 6,786, Level: 54
    Level completed: 18%, Points required for next Level: 164

    Location
    Sweden
    Posts
    524
    Thanks
    44
    Thanked 112 Times in 100 Posts

    Re: Determining the weight of variables

    If there wouldn't have been any variation in genes between humans then the coefficient of variation between genes and body length would be zero. Albeit this, length would still be determined to a great extent by genes.

  4. #4
    Points: 8, Level: 1
    Level completed: 15%, Points required for next Level: 42

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Determining the weight of variables

    Englund,
    Thanks. I am not looking for an exact number. In your example where you are giving genes a 90% weight, where are you pulling that estimate from?

    Mike

  5. #5
    Points: 3,666, Level: 38
    Level completed: 11%, Points required for next Level: 134

    Posts
    48
    Thanks
    3
    Thanked 5 Times in 5 Posts

    Re: Determining the weight of variables


    Keep your eye out for colinearity between your "independant" variables. The more correlated the independant variables are, the less robust this sort of interpriatation is.

+ Reply to Thread

           




Tags for this 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