2
$\begingroup$

I'm doing an experiment with moodcongruency (influence of mood on word retrieval). Prior to inducing a mood, the participants will memorize a list of words containing an equal amount of positive, negative or neural words.

Before the actual experiment, I want to make sure that the words are indeed associated (by a sample) with a positive, negative or neutral mood. So I've summarized a list containing 37 words and asked 24 respondents to rate their association with either (1) negative, (2) neutral of (3) positive meaning (nominal).

From these results, I'm looking to distil the most reliable words into a list for each association (pos, neg, neut).

How would I perform this in SPSS?

$\endgroup$
2
  • $\begingroup$ Can you describe what your input and desired output would be, in a bit more detail? You might be able to use something like a simple ranking or contingency table and pick the n-best words. Also you may be better off using an existing set of valenced words, or something like www.liwc.net. $\endgroup$
    – paul
    Dec 7, 2013 at 20:37
  • $\begingroup$ My SPSS input is just the set of individual words, rated 24 times, each words rated either a 1 (negative association) 2 (neutral association) or 3 (positive association). Also the words are in Dutch. I've noticed is that most words have a certain degree of variance. Not all words are equally neutral. Is there a way to calculate if this variance is significant? $\endgroup$
    – Daniel
    Dec 8, 2013 at 10:12

1 Answer 1

1
$\begingroup$

You could use Cronbach's alpha. In SPSS it's under Analyze > Scale > Reliability Analysis. There's a lot of information about it on the web as it's a very common method of reliability analysis for test items, and the way you're using these words is basically as test or Likert-type items. Scale Development by DeVellis is a pretty good resource for more info.

You might also be interested in looking at the development of the LIWC system, which analyzes text for "positive or negative emotions, self-references, causal words, and 70 other language dimensions". The creators have many publications discussion their methods in dealing with a very similar problem.

$\endgroup$

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.