Question Regarding Overlapping Demographics

[Brian Dates]

Greetings, all. I'm running an ANOVA, analyzing changes in service delivery due to the implementation of telemental health. I have an overlap between Race and Ethnicity. Race is divided into Black/African American, White, and Other Race. Ethnicity is Latinx/Non-Latinx. 78.7% of the Other Race is Latinx. 70% of Latinx is Other Race. So the majority of Latinx respondents used Other Race. If I eliminate Latinx subjects from Other Race, the result is 298 subjects in the Other Race category compared to 4676 Black/African American and 4038 White. If I eliminate those 298 subjects from Other Race, then that effectively makes Hispanic as a racial category with 1101 members. So I might use four racial categories and eliminate ethnicity altogether knowing that one of the racial categories was Hispanic. Any thoughts?

Thanks in advance. I suspect I'm not alone in this quandary with respondent confusion regarding ethnicity and race.

B

 

===================== 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

[Art Kendall]

Of course a lot depends on the purposes of the study and the sampling design.

2 suggestions.
Assuming a simple random sample.
1 (easier) crosstab the 2 variables, create a third variable nominal
variable. If you have empty or tiny cell n's collapse some. You could have a
six value one-way anova.

2. go to the OMB website and see what the US Federal reporting categories
are.


Do you have the populations sizes for the area of interest?  Are the sample
sizes very inconsistent with the pop?

Are the cases actually the population?



-----
Art Kendall
Social Research Consultants
--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
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

[Rich Ulrich]

Who wants to know what, and why?

I think you will run into a lot of flak if you put Latinx out there as

a Race. The bulk of the sample is Black/white; the sample size is

large, considering the hypotheses.  That is, if the only "effects" are

small ones, you would not have much to point to, and you would

want to control for other variables that would account for small effects --

age, income, computer-access, family role, family size, and so on. 

I think I would start by focusing on the Black/white, and look at all

those side issues to see how much they matter.  That would instruct

me on how to further incorporate Latinx or Other race. Just displaying

certain sets of means might be enough, beyond the initial tests based

on Black/white.

--

Rich Ulrich

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of Brian Dates <[hidden email]>
Sent: Tuesday, January 12, 2021 12:48 PM
To: [hidden email] <[hidden email]>
Subject: Question Regarding Overlapping Demographics

 

Greetings, all. I'm running an ANOVA, analyzing changes in service delivery due to the implementation of telemental health. I have an overlap between Race and Ethnicity. Race is divided into Black/African American, White, and Other Race. Ethnicity is Latinx/Non-Latinx. 78.7% of the Other Race is Latinx. 70% of Latinx is Other Race. So the majority of Latinx respondents used Other Race. If I eliminate Latinx subjects from Other Race, the result is 298 subjects in the Other Race category compared to 4676 Black/African American and 4038 White. If I eliminate those 298 subjects from Other Race, then that effectively makes Hispanic as a racial category with 1101 members. So I might use four racial categories and eliminate ethnicity altogether knowing that one of the racial categories was Hispanic. Any thoughts?

Thanks in advance. I suspect I'm not alone in this quandary with respondent confusion regarding ethnicity and race.

B

 

===================== 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

===================== 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

[bdates]

Everyone,

Thank-you very much! The population is all the community mental health consumers in the Detroit-Wayne County system. The sample is purposive cluster to assure geographic representativeness, which means including various racial/ethnic groups. The Other Race group is comprised of those subjects who answered "Other Race" and the very small individual racial groups not Black or White. The latter groups were identified by race but none of the individual categories amounted to very many so they were included in the Other Race group. Additionally, the categories themselves contain heterogeneous cultures. For example, the largest of them is Asian. This includes Japanese, Chinese, and Hmong - clearly very different cultures. So in keeping with the input, I'm keeping Ethnicity and reducing Race to Black/White. Looking at differences between Black, White, and Other is rather meaningless since Other is both very small in comparison of N's, but also heterogeneous. So it becomes nothing but a placeholder, and a significant finding for Other doesn't really tell me anything.

Make sense?

Brian

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of Rich Ulrich <[hidden email]>
Sent: Tuesday, January 12, 2021 3:33 PM
To: [hidden email] <[hidden email]>
Subject: Re: Question Regarding Overlapping Demographics

 

Who wants to know what, and why?

I think you will run into a lot of flak if you put Latinx out there as

a Race. The bulk of the sample is Black/white; the sample size is

large, considering the hypotheses.  That is, if the only "effects" are

small ones, you would not have much to point to, and you would

want to control for other variables that would account for small effects --

age, income, computer-access, family role, family size, and so on. 

I think I would start by focusing on the Black/white, and look at all

those side issues to see how much they matter.  That would instruct

me on how to further incorporate Latinx or Other race. Just displaying

certain sets of means might be enough, beyond the initial tests based

on Black/white.

--

Rich Ulrich

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of Brian Dates <[hidden email]>
Sent: Tuesday, January 12, 2021 12:48 PM
To: [hidden email] <[hidden email]>
Subject: Question Regarding Overlapping Demographics

 

Greetings, all. I'm running an ANOVA, analyzing changes in service delivery due to the implementation of telemental health. I have an overlap between Race and Ethnicity. Race is divided into Black/African American, White, and Other Race. Ethnicity is Latinx/Non-Latinx. 78.7% of the Other Race is Latinx. 70% of Latinx is Other Race. So the majority of Latinx respondents used Other Race. If I eliminate Latinx subjects from Other Race, the result is 298 subjects in the Other Race category compared to 4676 Black/African American and 4038 White. If I eliminate those 298 subjects from Other Race, then that effectively makes Hispanic as a racial category with 1101 members. So I might use four racial categories and eliminate ethnicity altogether knowing that one of the racial categories was Hispanic. Any thoughts?

Thanks in advance. I suspect I'm not alone in this quandary with respondent confusion regarding ethnicity and race.

B

 

===================== 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

===================== 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

===================== 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