f



Advice on Type of Analysis

Hello SPSS experts!

I was just struggling with what kind analysis I might carry out for a certa=
in dataset.=20

I have a sample of 200 cases (IDs) divided into three different groups; I w=
ould like to analyze the scores of each intervention (score having taken a =
pre-assessment, score taking some pre-assessment, and score with no pre-ass=
essment). Initially, I was thinking of using an ANOVA, however the dataset =
violates more than a few assumptions (e.g., the subjects have duplicate sco=
res, some subjects are in all three groups multiple times, no homoscedastic=
ity, data were not randomly sampled).=20

So since I just want to analyze whether having a degree intervention influe=
nces scores, I figure an ANOVA would be appropriate, and I would not be loo=
king for between-subject/group effects.

Would I be okay in carrying out this kind of analysis if I'm simply looking=
 at whether having full/some/no pre-assessment has an overall impact on the=
ir final grade in a class?=20

Thank you in advance for the help!
0
jace
11/6/2016 10:16:40 PM
comp.soft-sys.stat.spss 5679 articles. 0 followers. Post Follow

3 Replies
458 Views

Similar Articles

[PageSpeed] 20

On Sun, 6 Nov 2016 14:16:40 -0800 (PST), jace.riley@gmail.com wrote:

>Hello SPSS experts!
>
>I was just struggling with what kind analysis I might carry out for a certain dataset. 

I think you will have to describe your design with less ambiguity ....

>
>I have a sample of 200 cases (IDs) divided into three different groups; I would like to analyze the scores of each intervention (score having taken a pre-assessment, score taking some pre-assessment, and score with no pre-assessment). 

Does this describe a /battery/ of tests at Pre?  - I read it as this:
some people took the whole battery, some took part of it, and 
some took none of it.  Are these three groups being described as
three different "interventions"?  (Is your outcome score something
that is also being measured at Pre?)

>               Initially, I was thinking of using an ANOVA, however the dataset violates more than a few assumptions (e.g., the subjects have duplicate scores, some subjects are in all three groups multiple times, no homoscedasticity, data were not randomly sampled). 

Totally unclear.  Why are there duplicate scores?
My interpretation, above, does not allow for "some subjects" to be 
in "all groups" even once, not to mention "multiple times".  How?

If there is no homoscedasticity -- Why?  How does that happen?
Are your outcome scores badly scaled?  Are you observing strong
effects so that (say) a lot of subjects score near 100% at the end? 

>
>So since I just want to analyze whether having a degree intervention influences scores, I figure an ANOVA would be appropriate, and I would not be looking for between-subject/group effects.

The absence of randomization means that you should not place any
faith in p-values unless you can argue for the expectation of "all 
things being equal".   

>
>Would I be okay in carrying out this kind of analysis if I'm simply looking at whether having full/some/no pre-assessment has an overall impact on their final grade in a class? 

"association" seems like a better, more hedge-full term, than
"impact".   But, as I describe above, my interpretation of your design
has multiple contradictions.  

-- 
Rich Ulrich 

0
Rich
11/7/2016 12:38:40 AM
On Sunday, November 6, 2016 at 7:38:46 PM UTC-5, Rich Ulrich wrote:
> On Sun, 6 Nov 2016 14:16:40 -0800 (PST), jace.riley@gmail.com wrote:
>=20
> >Hello SPSS experts!
> >
> >I was just struggling with what kind analysis I might carry out for a ce=
rtain dataset.=20
>=20
> I think you will have to describe your design with less ambiguity ....
>=20
> >
> >I have a sample of 200 cases (IDs) divided into three different groups; =
I would like to analyze the scores of each intervention (score having taken=
 a pre-assessment, score taking some pre-assessment, and score with no pre-=
assessment).=20
>=20
> Does this describe a /battery/ of tests at Pre?  - I read it as this:
> some people took the whole battery, some took part of it, and=20
> some took none of it.  Are these three groups being described as
> three different "interventions"?  (Is your outcome score something
> that is also being measured at Pre?)
>=20
> >               Initially, I was thinking of using an ANOVA, however the =
dataset violates more than a few assumptions (e.g., the subjects have dupli=
cate scores, some subjects are in all three groups multiple times, no homos=
cedasticity, data were not randomly sampled).=20
>=20
> Totally unclear.  Why are there duplicate scores?
> My interpretation, above, does not allow for "some subjects" to be=20
> in "all groups" even once, not to mention "multiple times".  How?
>=20
> If there is no homoscedasticity -- Why?  How does that happen?
> Are your outcome scores badly scaled?  Are you observing strong
> effects so that (say) a lot of subjects score near 100% at the end?=20
>=20
> >
> >So since I just want to analyze whether having a degree intervention inf=
luences scores, I figure an ANOVA would be appropriate, and I would not be =
looking for between-subject/group effects.
>=20
> The absence of randomization means that you should not place any
> faith in p-values unless you can argue for the expectation of "all=20
> things being equal".  =20
>=20
> >
> >Would I be okay in carrying out this kind of analysis if I'm simply look=
ing at whether having full/some/no pre-assessment has an overall impact on =
their final grade in a class?=20
>=20
> "association" seems like a better, more hedge-full term, than
> "impact".   But, as I describe above, my interpretation of your design
> has multiple contradictions. =20
>=20
> --=20
> Rich Ulrich

Hi Rich,=20

Thanks for the response, I know it's quite an unconventional data set for S=
PSS analysis. I am attempting to determine whether a full pre-assessment, a=
 partial pre-assessment or no pre-assessment for certain math courses has a=
n effect on the overall course grade of a student. One student could have h=
ad a pre-assessment for one MAT course, but only a partial assessment for a=
nother, which I've labeled groups 1, 2 and 3. Each group has a correspondin=
g final course grade, but no score for the pre-assessments, as they are pas=
s/fail.

For example, ID #1 will have taken a pre assessment for one math course (MA=
T 01), but not for MAT 2, and perhaps has only completed a partial assessme=
nt for MAT 3, thus being included in all 3 groups. Again, this doesn't real=
ly involve a between groups analysis, but simply whether being given a pre-=
assessment has any effect on students' final course grade.=20

Hope this clears things up. Thanks again for the help.
0
darko
11/7/2016 12:56:23 AM
On Sun, 6 Nov 2016 16:56:23 -0800 (PST), darko.giacomini@gmail.com
wrote:

>On Sunday, November 6, 2016 at 7:38:46 PM UTC-5, Rich Ulrich wrote:
>> On Sun, 6 Nov 2016 14:16:40 -0800 (PST), jace.riley@gmail.com wrote:
>> 
>> >Hello SPSS experts!
>> >
>> >I was just struggling with what kind analysis I might carry out for a certain dataset. 
>> 
>> I think you will have to describe your design with less ambiguity ....
>> 
>> >
>> >I have a sample of 200 cases (IDs) divided into three different groups; I would like to analyze the scores of each intervention (score having taken a pre-assessment, score taking some pre-assessment, and score with no pre-assessment). 
>> 
>> Does this describe a /battery/ of tests at Pre?  - I read it as this:
>> some people took the whole battery, some took part of it, and 
>> some took none of it.  Are these three groups being described as
>> three different "interventions"?  (Is your outcome score something
>> that is also being measured at Pre?)
>> 
>> >               Initially, I was thinking of using an ANOVA, however the dataset violates more than a few assumptions (e.g., the subjects have duplicate scores, some subjects are in all three groups multiple times, no homoscedasticity, data were not randomly sampled). 
>> 
>> Totally unclear.  Why are there duplicate scores?
>> My interpretation, above, does not allow for "some subjects" to be 
>> in "all groups" even once, not to mention "multiple times".  How?
>> 
>> If there is no homoscedasticity -- Why?  How does that happen?
>> Are your outcome scores badly scaled?  Are you observing strong
>> effects so that (say) a lot of subjects score near 100% at the end? 
>> 
>> >
>> >So since I just want to analyze whether having a degree intervention influences scores, I figure an ANOVA would be appropriate, and I would not be looking for between-subject/group effects.
>> 
>> The absence of randomization means that you should not place any
>> faith in p-values unless you can argue for the expectation of "all 
>> things being equal".   
>> 
>> >
>> >Would I be okay in carrying out this kind of analysis if I'm simply looking at whether having full/some/no pre-assessment has an overall impact on their final grade in a class? 
>> 
>> "association" seems like a better, more hedge-full term, than
>> "impact".   But, as I describe above, my interpretation of your design
>> has multiple contradictions.  
>> 
>> -- 
>> Rich Ulrich
>
>Hi Rich, 
>
>Thanks for the response, I know it's quite an unconventional data set for SPSS analysis. I am attempting to determine whether a full pre-assessment, a partial pre-assessment or no pre-assessment for certain math courses has an effect on the overall course grade of a student. One student could have had a pre-assessment for one MAT course, but only a partial assessment for another, which I've labeled groups 1, 2 and 3. Each group has a corresponding final course grade, but no score for the pre-assessments, as they are pass/fail.
>
>For example, ID #1 will have taken a pre assessment for one math course (MAT 01), but not for MAT 2, and perhaps has only completed a partial assessment for MAT 3, thus being included in all 3 groups. Again, this doesn't really involve a between groups analysis, but simply whether being given a pre-assessment has any effect on students' final course grade. 
>
>Hope this clears things up. Thanks again for the help.

This description is fairly self-consistent, but it is incomplete;
and it contradicts the original in an important aspect, namely,
one ID can belong to each "group" for a different course.

Now it seems to be:

There are 200 IDs.  Each ID has one or more math classes, 
out of some unstated number of courses. 

For each ID and class, there may be a Pre, PartialPre, or NoPre;
each of Pre and PartialPre is scored only as Pass/Fail.  (? yes?)
 - This might suggest examining scores for 5 groups, those being
NoPre and the Pass and Fail for each of the others.... either to see
what they look like, or as the main result of your data collection. 

There is no comment on WHO had a Pre, or whether anyone was
to know the result or to make use of it.  Was there ever a Pre=fail
followed by the person not-taking the course because they or the
teacher knew of the result? 

Was the Pre only administered to those whose success was 
deemed unlikely?  (Is this a test that is "administered" or is it
come other sort of assessment.  WHY is it only Pass-Fail?)

With "unbalanced" data, it does seem that you want to use
VarsToCases to put your data in the long form, so that you
potentially can use a Mixed analysis that controls for ID and
Course, with Assessment entered as one or two covariates.

-- 
Rich Ulrich 

0
Rich
11/8/2016 3:37:13 AM
Reply: