Query on model

[Muni Rubens]

Hello,

 

I'm a graduate student in public health.

 

Right now I'm analyzing HIV knowledge in a sample which consists of Hispanics and African Americans. We have a 37 item questionaire for HIV knowldege and I scored it and did an independent sample t-test to know whether there is a change in knowledge between African Americans and Hispanics. I found a difference at p=0.000 with Levene's at p=0.78.

 

Now I want to know which are the predictors of HIV knowledge. So I've taken demographics (age, gender, yeras of schooling, etc) and ethnic identity score (this is a score of 5 item questionaire with Likert scale 1 to 4) as my independent and HIV knowledge score as dependent variable and did a linear regression. I found that years of schooling and ethnic identity score to have a sig. predictive effect. Literature review shows that only years of schooling has sig. effect. There is no sig. correlation between years of schooling and ethnic identity score.

 

My question is whether the model I described above is good for the analysis I've done. My sample size is 390; 140 African Americans and 250 Hispanics. Is it OK to take the whole HIV score for African Americans and Hispanics and do linear regression to find my predictors ETHNIC IDENTITY or categorize knowledge and do a logistc to find the difference in level of knowledge.

--
Thanks,
Muni Rubens

"In every great act, there is a challenge. In every challenge, there is a reward. In every reward lies the product of our efforts. In every effort lies new beauty to be born."

[Maguin, Eugene]

Muni,

I'd wonder about two questions. The first you might well judge to be
irrelevant. It is whether those 37 items form a single well-structured,
unidimensional scale or two or more such scales. The second is about the
question you asked. If you were to dichotomize the HIV measure, what would
be your rationale for doing so, what would be the question you want to
answer? Is there, for instance, such a thing as a 'passing grade' that is
more useful to identify predictors for than simply a total score?

But, here is something that might be useful to look into. Is the
relationship between the IVs and DV linear? Are there interactions between
pairs of your IVs?

Gene Maguin



>>Right now I'm analyzing HIV knowledge in a sample which consists of
Hispanics and African Americans. We have a 37 item questionaire for HIV
knowldege and I scored it and did an independent sample t-test to know
whether there is a change in knowledge between African Americans and
Hispanics. I found a difference at p=0.000 with Levene's at p=0.78.

Now I want to know which are the predictors of HIV knowledge. So I've taken
demographics (age, gender, yeras of schooling, etc) and ethnic identity
score (this is a score of 5 item questionaire with Likert scale 1 to 4) as
my independent and HIV knowledge score as dependent variable and did a
linear regression. I found that years of schooling and ethnic identity score
to have a sig. predictive effect. Literature review shows that only years of
schooling has sig. effect. There is no sig. correlation between years of
schooling and ethnic identity score.

My question is whether the model I described above is good for the analysis
I've done. My sample size is 390; 140 African Americans and 250 Hispanics.
Is it OK to take the whole HIV score for African Americans and Hispanics and
do linear regression to find my predictors ETHNIC IDENTITY or categorize
knowledge and do a logistc to find the difference in level of knowledge.


--
Thanks,
Muni Rubens

"In every great act, there is a challenge. In every challenge, there is a
reward. In every reward lies the product of our efforts. In every effort
lies new beauty to be born."

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD