A question about proper statistical test for research on attitudes

Home Forums Methodspace discussion A question about proper statistical test for research on attitudes

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #788
    Marko Sobol
    Member

    Hi,

    I was asked for help in choosing adequate statistical test for data analysis, and now I am struggling with it too. Research in question resulted in two sets of data gathered by questionnaire targetting three different groups of people (divided by age), being asked about their attitude on the current state of the city they live in (first set of data) and after that their wished state about the city they are living in (second set of data). Answers were collected on 5 point Likert scale. Student wants to analyse if there are any statistically significant differences between the answers among the three groups (groups are divided by age, 20-39, 40-59, 60-79) as well as between the two attitudes – on the current state of the city and their wished state of the city. Student wants to do ANOVA, to compare answers between the three age groups as well the two attitudes (current and wished state). Would this be a within-subjects design (all age groups went through the same questions on questionnaire), or between-subjects design (groups are different, a person belonging to 20-39 age group cannot obviously belong to any other group), and what kind of ANOVA would be appropriate? 3X3 repeated measures ANOVA or 3×3 between subjects ANOVA, mixed ANOVA, MANOVA? Please help.    

    #793
    Dave Collingridge
    Participant

    ANOVA sounds like a reasonable design. Your next decision should be whether the ANOVA is univariate or multivariate. If you aggregate the responses for both sets of data and analyze the outcomes separately then you have a factorial design with one outcome, and 2-3 between subjects factors. If you want to analyze both aggregated outcomes at the same time then your looking at a multivariate design (MANOVA) with 2-3 between subjects factors. It looks like you have a between-subjects design. Within subjects designs typically come up when you have repeated measures (i.e., a variable that is similar across all participants like, “time’). 

    #792
    Marko Sobol
    Member

    Thank you very much, this helps. I’m still looking at the MANOVA or possibly Mixed ANOVA since different groups went through two different conditions or two dependent varibales, but most likely I will opt for One way MANOVA.

    #791
    Stephen Gorard
    Participant

    I would strongly advise not using any test of significance. Then the issue does not arise. 

    Instead simply report overall results for both ‘attitudes’, and again for cases divided into three age groups. What else is needed?

    #790
    Dr. james sika
    Participant

    ATTITUDE IS MEASURABLE. CANNOT BE IGNORED. USE OF ANOVA WILL DO

    DR.JAMES

    #789
    Stephen Gorard
    Participant

    Putting things in capitals does not make it any more valid. ANOVA does not work as intended, and it is not at all clear that the questioner has a complete random sample anyway. 

Viewing 6 posts - 1 through 6 (of 6 total)
  • You must be logged in to reply to this topic.