ANOVA and SPSS confusion

Home Forums Default Forum ANOVA and SPSS confusion

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
    Posts
  • #1572
    Martha
    Member

    Hello,

    I have just had to chage some elements of my dissertation as my hypothesis weren’t experimental but now I am stumped as to what to do.

    Let me explain my experiment so far;

    I have 4 conditions;

    Cond 1: High IV1, High IV2

    Cond 2: High IV1, Low IV2

    Cond 3: Low IV1 High IV2

    Cond 4: Low IV1 Low IV2

    My hypothesises are:

    Hypothesis 1: Participants in the High Iv1/ High iv2 condition will score higher on the         XYZ scale than participants in the Low IV/ Low IV2 condition.

    Hypothesis 2: Participants in the Low Iv1 / Low IV2 condition will score higher on the ZYX than participants in the High IV/ High IV2  condition.

    Hypothesis 3: Participants in the High IV1/ High IV2 condition will score higher on the XYZ scale than participants in the Low IV1/ High IV2 condition.

    Hypothesis 4: Participants in the High IV1/ High IV2 condition will score higher on the XYZ scale than participants in the High IV1/ Low IV2 condition.

    My issue is I have no idea how to test these hypotheses. I think I need to do a Univariate ANOVA but have no idea how to analyse the results based on conditions, if that makes sense?

    Any help would be much apprieciated, happy to clarify more if necessary.

    #1580
    Rafael Garcia
    Participant

    Martha,

    Based on this description, I have a question. Are you comparing Conditions 2-4 to Condition 1? If so, you could simply do dummy coding with Condition 1 as the reference group. 

    It looks like, however, that you’re trying to test for order effects. How is your data organized?

    Raf

    #1579
    Martha
    Member

    Hi,

    Thanks for your quick reply!

    Forgot to mention it is a 2×2 factorial between subjects design.

    I have downloaded my data straight from Qualtrics, what exactly do you mean by how is my data organised? Stats really isn’t my strong point – I am sorry, you will have to explain it very simply please!

    Basically my study is a questionnaire which involves ppts getting 1 of 4 vignettes (thus assigning them to a condition) and then having to answer 4 scales (relating to DV1, DV2 and manipulation check scales, relating to IV1 and IV2).

    #1578
    Rafael Garcia
    Participant

    So you have 2 levels of IV1 and two levels of IV2 and want to relate them to DV1 and DV2?

    #1577
    Martha
    Member

    Yes, the two DVs are negatively correlated, if that makes any difference.

    So I want to see if;

    High levels of both IVs together lead to significantly higher results on DV1 (condition 1) than in condition 4 (low levels of both Dvs)

    There will be a difference in results on the scale to measure DV1 between condition 2 and 3.

    Just High levels of IV1 lead to significantly higher results of DV1 than High levels of IV2. (i.e which IV is a stronger predictor of DV1)

    I am not sure if it will be possible to tell the last one from my design.

    Hope that clears things up a little more, once I know how to do the statistics fo DV1 I should be able to do it for DV2.

    Thanks again

    #1576
    Rafael Garcia
    Participant

    There are three approaches I can see:

    1) run this as a 4 group ANOVA and do post hoc tests to test each of the differences that you suspect. 

    2) dummy code (0,1) IV1 and IV2 and do a sequential/hierarchical regression putting IV1’s dummy code first.

    3) if you can express this in planned orthogonal contrasts, do contrast coding.

    Given your description, I’m leaning toward option 2. You can do a regression for each DV with the dummy codes for IV1, IV2, & the interaction. 

    #1575
    Martha
    Member

    Hi,

    Thanks for your reply.

    I had a look at dummy coding last night but I don’t really understand how I will code them.

    So if for example, I have IV1 as my baseline (and therefore assign it 0 – please correct me if I am wrong at all, dummy coding is a completly new topic for me), how would I include the different levels? Also, I read that you have one less dummy variable for each comparisiongroup, would a group be a condition? I don’t really understand having one less variable as then surely some factors get missed out?

    All conditions include both IVs, albeit different levels, so wouldn’t that affect the coding?

    #1574
    Rafael Garcia
    Participant

    Create a variable called D1, you will have LowIV1 = 0, HighIV1 = 1.

    Create a second called D2, you will have LowIV2 = 0, HighIV2 = 1.

    Looking at the unstandardized regression coefficients:

    The intercept will be the mean for Cond4. D1 will tell you the added effect of having High IV1. D2 will be the added effect of having High IV2.  

    Just High levels of IV1 lead to significantly higher results of DV1 than High levels of IV2. (i.e which IV is a stronger predictor of DV1)

    If this is true, then D1 will be significant only when predicting DV1 and D2 will be significant only when predicting DV2.

    High levels of both IVs together lead to significantly higher results on DV1 (condition 1) than in condition 4 (low levels of both Dvs)

    This will be true if either D1 or D2 are significantly different than zero.

    There will be a difference in results on the scale to measure DV1 between condition 2 and 3.

    You can explore this more with post hocs if the interaction (D1*D2) is significant. 

    #1573
    Martha
    Member

    Hi,

    I have dummy coded and did a Univariate analysis (no idea if this was the right thing to do) with D1 and D2 as my IVs and DV1 as my dv. In the interaction bit (d1*d2) it was left blank, and I got unsignificant results (.673 at .05 level) for d2, which was suprising. Have I done this right? Or do I need to do a diff analysis?

    Thank you!

Viewing 9 posts - 1 through 9 (of 9 total)
  • The forum ‘Default Forum’ is closed to new topics and replies.