Testing covariables ( doing an(c)ova with multiple independent variables?)? in SPSS

esbag

New Member
#1
Hi everyone,

I'm stuck in my analyses of my covariables.
Here are the questions I want to test:
What is the influence of educational level on judgments and decisions of participants?​
What is the influence of organizational level on judgments and decisions of participants?​
What is the influence of emotional intelligence on judgments and decisions of participants?​
What is the influence of social / time pressure on judgments and decisions of participants?​

I thought of doing an AN(C)OVA.

Dependent variable =
applicant credential (or applicant trait importance/ applicant total evaluation (separate analyses for the three different dependent variables).

Categorical independent variable =
candidate (western/non western)
experimental condition (both, ethnic prime, person prime, control group)
(Time/social pressure; Is either 1 or 2 (yes and no time pressure (but is this an independent variable or a covariable?))

Covariables =
Educational level (continuous)
Organization level (continuous)
Total emotional intelligence score (continuous) (tot_tei)


This is what my syntax has to look like according to my stats teacher :
( my statistics teacher told me I had to put in contrasts (I don’t know what for?? And he put all of the covariables in once)).

UNIANOVA appl_credent BY candidate exp_condition time_pressure WITH edu_level org_level tot_tei
/CONTRAST(candidate)=Polynomial
/CONTRAST(exp_condition)=Helmert
/CONTRAST(time_pressure)=Simple
/METHOD=SSTYPE(3)/INTERCEPT=INCLUDE
/PLOT=PROFILE(exp_condition*candidate*time_pressure) TYPE=LINE ERRORBAR=CI MEANREFERENCE=NOYAXIS=0
/PRINT ETASQ DESCRIPTIVE PARAMETER OPOWER/CRITERIA=ALPHA(.05)
/DESIGN=edu_level org_level tot_tei candidate exp_condition time_pressurecandidate*exp_condition candidate*time_pressure exp_condition*time_pressurecandidate*exp_condition*time_pressure.

Now I am not sure if I have to
  1. Put all the covariables in the analyses all in once (or first without covariables and then one by one and the last analyses only the signifanct covariables altogether)
  2. Put “contrasts” in the analyses, since I don’t know what to use that for..
  3. Where to look in the output. If I want to find answers for my control variable questions: (e.g. What is the influence of educational level /organizational level /emotional intelligence/ time pressure on judgments and decisions of participants?

I sincerely hope my question and explanation is clear.I’m stuck and I would be so grateful for your help and opinions!!
 

Karabiner

TS Contributor
#2
What is the influence of educational level on judgments and decisions of participants?
Really, just the main effects of these variables? Not their interaction with candidate type and/or
experimental condition, i.e. moderator effects (the interactions will be computed automatically
by SPSS, if I remember correctely)?

Educational level (continuous)
Organization level (continuous)
But these are ordinal scaled variables!?

my statistics teacher told me I had to put in contrasts (I don’t know what for?.
Candidate and time_pressure each have only 2 levels, so contrasts are not needed there, indeed.
And he put all of the covariables in once.
Why shouldn't he?

How large is your sample size, by the way?

Where to look in the output.
Not sure what this question precisely means. In the ANOVA output you will see the names of the
respective variables and the tests and results associated with them.

With kind regards

Karabiner
 

esbag

New Member
#3
Hi,

Thanks for your reply!

Really, just the main effects of these variables? Not their interaction with candidate type and/or
experimental condition, i.e. moderator effects (the interactions will be computed automatically
by SPSS, if I remember correctely)?
I have 16 groups which can be divided into four groups: e.g.: both primes, ethnic prime only, personal objectivity prime only, control groups (no prime at all). (See attachment).

Within these groups participants either got a western candidate (variable candidate =1) or a non-western candidate (variable candidate =2). They had to rate this candidate (applicant credential rating, trait importance rating and total evaluation).

For my first hypotheses (see below) I did the following:
1. Participants primed with ethnic prejudice rate the credentials of the Western applicant higher than the same credentials of the non-Western applicant.
2. Participants primed with personal objectivity rate the credentials of the Western applicant higher than the same credentials of the non-Western applicant.

(And comparable hypotheses for applicant trait importance rating and total evaluation).

Here is what I did:

I split the file on the prime conditions (and only looked at the t-test results for ethnic prime =1, for example (so looked at people who where in that condition)). I did this for all four conditions. And put the results in a table.

SORT CASES BY ethn_Prime.
SPLIT FILE SEPARATE BY ethn_Prime.

T-TEST GROUPS=candidate(1 2)
/MISSING=ANALYSIS/VARIABLES=appl_credent appl_traitimport appl_totevaluation
/CRITERIA=CI(.95).

SPLIT FILE OFF.
SORT CASES BY persob_prime.
SPLIT FILE SEPARATE BY persob_prime.

T-TEST GROUPS=candidate(1 2)
/MISSING=ANALYSIS/VARIABLES=appl_credent appl_traitimport appl_totevaluation
/CRITERIA=CI(.95).

SPLIT FILE OFF.
etc..etc..


So to answer your question: "Really, just the main effects of these variables? Not their interaction with candidate type and/or
experimental condition, i.e. moderator effects (the interactions will be computed automatically
by SPSS, if I remember correctely)".
--> I don't know. In the ANCOVA syntax you can see that there are interactions if that's what you mean. I just want to look at the effect of educational level (and the other covariables) on the ratings (applicant credential, trait importance and total evaluation). That's also why I'm asking what to look for in the output of my analysis. do I report the interaction effect or the main effects etc.


But these are ordinal scaled variables!?
Yeah, as far as I know the covariates need to be continuous.. or am I saying something stupid?

Candidate and time_pressure each have only 2 levels, so contrasts are not needed there, indeed.
and for experimental condition?

Why shouldn't he?

How large is your sample size, by the way?
I don't know, I thought maybe I have to put them in one by one and check their effect (because the F- value is different if I put the covariables seperately or alltogether (see attachments). I am honestly so confused on how to interpret the output.

my sample size is N= 149

Not sure what this question precisely means. In the ANOVA output you will see the names of the
respective variables and the tests and results associated with them.
Yeah I know, but in the output I just don't know what I have to report.


With kind regards

Karabiner

Thanks for all your help!

Best,
 

Attachments

Karabiner

TS Contributor
#4
As far as I can see from the ANOVA table , experimental conditon, time pressure, stimulus,
2 out of 3 covariates, and the interactions are not statistically significantly associated with
appl_credent. The only statistically significant predictor is emotional intelligence. The
results seem quite clear, at least for this dependent variable.

With kind regards

Karabiner
 
#5
As far as I can see from the ANOVA table , experimental conditon, time pressure, stimulus,
2 out of 3 covariates, and the interactions are not statistically significantly associated with
appl_credent. The only statistically significant predictor is emotional intelligence. The
results seem quite clear, at least for this dependent variable.

With kind regards

Karabiner
Yes, that's true, but which results do I report in my "results section". Do I report the F-value and significance of educational level only if I want to answer my question? Or do I report more? (like also the interaction ?)

And as you can see the values are different, depending if I put all the covariables at once in the analyses (or one by one). So which result do I report, the analyses in which I put all of the covariables in once?

so for the question:
"What is the influence of educational level on judgments and decisions of participants?"

I write the following results (all covariables in the analyses all together) (middle attachment):
F value of 3.14 and significance of .078 for educational level on applicant credential rating?

Thank you so much for all your help btw!
 

Karabiner

TS Contributor
#6
which results do I report in my "results section". Do I report the F-value and significance of educational level only if I want to answer my question? Or do I report more? (like also the interaction ?)
How should I know? Every journal/department/professor may have its own standards for that.
Why can't you ask there, or at least have a look on what others have done there when presenting
their results? In psychology, for example, the APA standards could be used.
And as you can see the values are different, depending if I put all the covariables at once in the analyses (or one by one). So which result do I report, the analyses in which I put all of the covariables in once?
It depends on your theoretical background and on your precise research question(s).
Including the variables all at once would seem natural to me, since it analyses the
effect of one variable while the effects of other, probably overlapping variables are
also taken into account.

With kind regards

Karabiner