1
$\begingroup$

Can I compare outcome data from different literacy measures with different scales (delivered to different subjects) by transforming the data into z scores? Ideally I'd like to use regression and ANOVA techniques to determine if there are significant differences between groups that got different types of literacy interventions. Is this remotely plausible? If not, is there another legitimate way to compare the scores from different measures?

$\endgroup$
2
  • $\begingroup$ It sounds like you have multiple outcome measures, do I have that right? Also, I'm guessing that you suspect these outcomes are not independent of each other; eg, you suspect that they are all measures of the same underlying reality (ie, greater or lesser degrees of 'literacy'), is that right? $\endgroup$ Commented Feb 2, 2013 at 19:40
  • $\begingroup$ Yes, right. Multiple outcome measures -- although I'm working hard to narrow down the options. And yes, they all measure early literacy constructs. $\endgroup$
    – Margaret
    Commented Feb 2, 2013 at 21:10

1 Answer 1

4
$\begingroup$

The answer will depend on the following questions:

  1. How many times was each measure given to each subject?
  2. Were the subjects randomly assigned to the different interventions?
  3. Are the measures normed on some other population?

If, as I suspect, the answers are "1", "No" and "No" then I don't see how this can be done.

However, if the answer to 1 is "more than once" then you can compare how each subject changed over time.

If the answer to 3 is "yes" then you can use those norms.

If, in addition to either of these, the answer to 2 is "yes" then you can generalize.

But it would help us if you gave more information; you may find this post from my blog helpful: How to ask a statistics question.

$\endgroup$
4
  • $\begingroup$ 1. More than once -- this is a multi-year initiative. $\endgroup$
    – Margaret
    Commented Feb 2, 2013 at 21:11
  • $\begingroup$ 2. The participants aren't randomly assigned in most cases, but are using Propensity Score Matching. And 3. Yes, they have been normed. How would you use the norms in this case? $\endgroup$
    – Margaret
    Commented Feb 2, 2013 at 21:16
  • $\begingroup$ If the measures have been normed, you can then use z on the normed data to compare (as long as both measures were normed on similar populations). That is, don't calculate z within each sample, use mean and sd of the normed group to create a z. $\endgroup$
    – Peter Flom
    Commented Feb 2, 2013 at 22:42
  • 1
    $\begingroup$ Can you provide an academic reverence for your answer regarding the possibility to perform repeated measures anova on different scales by using z-scores? Thanks in advance; Tomer $\endgroup$ Commented Sep 15, 2013 at 10:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.