Help with weighting items to create a new factor

Discussion in 'SPSS' started by Laura Torissi, Oct 15, 2011.

  1. Hello group- I am attempting to write syntax to create a new scale
    score for my data-you will see below that there are four scale scores.
    I would like to negatively weight the items that compose the scales
    Helpless and Unhealthy and Meat Preference and add to that the items
    of the scales Health Consciousness and Food Exploration, to result in
    an overall scale score called Healthy attitude total. Any guidance/
    advice on exactly how to do this?

    The negatively weighted items: nas2 nas4 nas5 nas9 nas10 nas15
    nas16 nas17 nas18 nas19 nas1 nas7 nas8 nas23

    Items as is: nas6 nas14 nas21 nas22 nas24 nas11 nas12 nas13

    (Below is the syntax for the scales, fyi)
    reliability variables = nas2 nas4 nas5 nas9 nas10 nas15 nas16
    nas17 nas18 nas19/scale(nashu) = nas2 to nas19.
    COMPUTE nashu =nas2 + nas4 + nas5 + nas9 + nas10 + nas15 + nas16 +
    nas17 + nas18 + nas19 .
    VARIABLE LABELS nashu 'NAS Helpless and Unhealthy scale score'.
    EXECUTE .

    reliability variables = nas6 nas14 nas21 nas22 nas24/scale(nasfe)
    = nas6 to nas24.
    COMPUTE nasfe =nas6 + nas14 + nas21 + nas22 + nas24.
    VARIABLE LABELS nasfe 'NAS Food Exploration scale score' .
    EXECUTE .

    reliability variables = nas1 nas7 nas8 nas23/scale(nasmp) = nas1 to
    nas23.
    COMPUTE nasmp =nas1 + nas7 + nas8 + nas23.
    VARIABLE LABELS nasmp 'NAS Meat Preference scale score' .
    EXECUTE .

    reliability variables=nas11 nas12 nas13/scale(nashc) = nas11 to nas13.
    COMPUTE nashc =nas11 + nas12 + nas13.
    VARIABLE LABELS nashc 'NAS Health Consciousness scale score' .
    EXECUTE .

    descriptives nashu to nashc.

    corr nashu to nashc.
     
    Laura Torissi, Oct 15, 2011
    #1
    1. Advertisements

  2. Laura Torissi

    Art Kendall Guest

    Why do you want to do this?
    There is a slight possibility that you can create a scale but it would
    not be a factor. A factor is a derived dimension obtained via factor
    analysis.

    What is the response scale for the items?

    How did you arrive at the number of scales and their labels?

    Do you have enough cases to allow you to factor all of the items at once?

    Are there items in each scale that are balanced for direction?

    Art Kendall
    Social Research Consultants
     
    Art Kendall, Oct 16, 2011
    #2
    1. Advertisements

  3. Laura Torissi

    Rich Ulrich Guest

    [snip, rest]

    To reverse a single item that runs from 1 to 4, you subtract from 5.
    Adjust as needed for your data, if you want to procede that way.

    You have scales that have different numbers of items. That, for
    me, has been a strong justification for using "Average" rather
    than "SUM" for my scaled scores. The other big reason is that
    "Average" preserves the tie to the original scale anchors. (A minor
    reason is that Average is usually exactly how I want to accommodate
    the existence of missing values.)

    To reverse-score a scale stated as Average, do as you would for items.

    To reverse-score a scale that is stated as Total, you can subtract
    from k (items) times the single-item reverser.

    With 4 sub-scales to combine, you have two obvious choices --
    weight each sub-scale equally, or weight each item equally.
    As a complication: "equal weight by sub-scale" probably means
    that you take into account the standard deviations.

    For equal-weight by item, you either add/subtract Totals,
    along with the reverser constant; or you multiple Averages
    to get totals, and do the same.

    To equal-weight by subscale, to take into account the standard
    deviations for an equal weighting, you abandon the item-labels
    as anchors. In this case, which is what I prefer, I
    eventually settled on presenting my composite scores with a
    mean of 50 and SD of 10.

    Thus, starting with Factors F1 to F4 and their standard deviations
    sd1 to sd4, needing to reverse F3 and F4, I would compute
    Temp= F1/sd1 + F2/sd2 - F3/sd3 - F4/sd4 .

    Then I would standardize Temp directly to (50,10), using
    the mean and SD as observed for Temp (sometimes for a
    baseline period alone, or else for the whole sample). That is,
    COMPUTE Composite= ( (Temp-meanTemp)/sdTemp ) * 10 + 50.0.
     
    Rich Ulrich, Oct 16, 2011
    #3
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.