SPSSX Discussion

Not sure whether nominal by interval test is appropriate

Classic

List

Threaded

11 messages Options

Mike Donatello

Not sure whether nominal by interval test is appropriate

All,

I’m having a difficult time with what should be a simple analysis. I don’t know whether I’m over-thinking the problem, but I can’t seem to figure out the appropriate method and syntax.

I have (T=11) different topic areas, each of which has K different potential sources of information. (The value of K varies among each of the 11 topic areas.) That is, for each {T=1 … 11} topic(s) in which a respondent is active there is the potential to have used {1 … K} different information sources. Also, for the most frequently used information source in each topic, the respondent provides a rating of quality.

What I am trying to find out -- within each single topic area in which a respondent is active, and across all topic areas taken together -- is whether there is a significant relationship between the quality rating and an information source being reported as the most frequently used.

For example, say I am active in topics A, B and C. For topic A, I used three sources; for topic B, I used nine sources; and for topic C, I used two sources. I list my most-used source for A, B and C, along with those sources’ respective quality ratings. How do I know whether the relationship between quality rating and a source being listed as most-used is (1) significant for topic A, B or C, and (2) significant regardless of topic?

Initially, I was thinking that this was a simple case of nominal (binary score for most-used or not) by interval (quality rating) correlation, but now I’m not sure whether that’s the proper approach. I have a nagging feeling that I am discarding useful information by going in that direction. (My dataset is structured in a way that will make that test a bear, but I think I can pull it off.) I assume that either logistic regression or point biserial correlation would be appropriate in that case.

Is there some other method that I am overlooking?

Thanks in advance for any help.

---

Mike Donatello

703.582.5680
[hidden email]

Mike Donatello

Re: Not sure whether nominal by interval test is appropriate

I should clarify that I do not have quality ratings for information sources
which are not most-used. If I did, the problem would be simpler. (At least
in my mind.)

MD

On Thu, 5 Apr 2012 20:41:44 -0400, Mike Donatello <[hidden email]> wrote:

>All,
>
>
>
>I'm having a difficult time with what should be a simple analysis. I don't
>know whether I'm over-thinking the problem, but I can't seem to figure out
>the appropriate method and syntax.
>
>
>
>I have (T=11) different topic areas, each of which has K different potential
>sources of information. (The value of K varies among each of the 11 topic
>areas.) That is, for each {T=1 . 11} topic(s) in which a respondent is
>active there is the potential to have used {1 . K} different information
>sources. Also, for the most frequently used information source in each
>topic, the respondent provides a rating of quality.
>
>
>
>What I am trying to find out -- within each single topic area in which a
>respondent is active, and across all topic areas taken together -- is
>whether there is a significant relationship between the quality rating and
>an information source being reported as the most frequently used.
>
>
>
>For example, say I am active in topics A, B and C. For topic A, I used three
>sources; for topic B, I used nine sources; and for topic C, I used two
>sources. I list my most-used source for A, B and C, along with those
>sources' respective quality ratings. How do I know whether the relationship
>between quality rating and a source being listed as most-used is (1)
>significant for topic A, B or C, and (2) significant regardless of topic?
>
>
>
>Initially, I was thinking that this was a simple case of nominal (binary
>score for most-used or not) by interval (quality rating) correlation, but
>now I'm not sure whether that's the proper approach. I have a nagging
>feeling that I am discarding useful information by going in that direction.
>(My dataset is structured in a way that will make that test a bear, but I
>think I can pull it off.) I assume that either logistic regression or point
>biserial correlation would be appropriate in that case.
>
>
>
>Is there some other method that I am overlooking?
>
>
>
>Thanks in advance for any help.
>
>
>
>---
>
>Mike Donatello
>
>703.582.5680
> <mailto:[hidden email]> [hidden email]
>
>
>
>
>
>
>
>

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

William Adams

Show only loess lines

In reply to this post by Mike Donatello

Hello Everyone,

Is it possible to show only smooth loess curves in a scatterplot? Basically, how can I hide or delete all the scatter dots and show only the loess lines?

Thank you!

Bill

Art Kendall

Re: Show only loess lines

With the graph editor for the output window, you should be able to select the marker(s) for the points and either delete them of change them to no color at all.

Art Kendall
Social Research Consultants

On 4/6/2012 11:02 AM, William Adams wrote:

Hello Everyone,

Is it possible to show only smooth loess curves in a scatterplot? Basically, how can I hide or delete all the scatter dots and show only the loess lines?

Thank you!

Bill

===================== 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
Social Research Consultants

ViAnn Beadle

Re: Show only loess lines

You can also generate a chart showing only the loess lines using the GGRAPH command with GPL. Here’s an example assuming variables x and y:

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=x y MISSING=LISTWISE

REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: x=col(source(s), name("x"))

DATA: y=col(source(s), name("y"))

GUIDE: axis(dim(1), label("x label"))

GUIDE: axis(dim(2), label("y label"))

ELEMENT: line(position(smooth.loess(x*y)), missing.wings())

END GPL.

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Art Kendall
Sent: Friday, April 06, 2012 9:15 AM
To: [hidden email]
Subject: Re: Show only loess lines

With the graph editor for the output window, you should be able to select the marker(s) for the points and either delete them of change them to no color at all.

Art Kendall

Social Research Consultants

On 4/6/2012 11:02 AM, William Adams wrote:

Hello Everyone,

Is it possible to show only smooth loess curves in a scatterplot? Basically, how can I hide or delete all the scatter dots and show only the loess lines?

Thank you!

Bill

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

ViAnn Beadle

Re: Show only loess lines

Sorry, I didn’t see any references to groups in your original post. Just add a color(RIDRETH1) function to your ELEMENT statement.

From: William Adams [mailto:[hidden email]]
Sent: Friday, April 06, 2012 9:54 AM
To: 'ViAnn Beadle'
Subject: RE: Show only loess lines

Hi ViAnn,

Thank you! Unfortunately, using the code below I am only seeing one line, but I need to see the loess lines for all of my groups. What am I missing?

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=RIDAGEYR MEAN(LBXSATSI)[name="MEAN_LBXSATSI"]

RIDRETH1 MISSING=LISTWISE REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: RIDAGEYR=col(source(s), name("RIDAGEYR"))

DATA: MEAN_LBXSATSI=col(source(s), name("MEAN_LBXSATSI"))

DATA: RIDRETH1=col(source(s), name("RIDRETH1"),

notIn("5"), unit.category())

GUIDE: axis(dim(1), label("Age"))

GUIDE: axis(dim(2), label("Mean ALT (U/L)"))

GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Race/Ethnicity - Recode"))

SCALE: cat(aesthetic(aesthetic.color.exterior), include("1", "2", "4", "6"))

ELEMENT: line(position(smooth.loess(RIDAGEYR*MEAN_LBXSATSI)), missing.wings())

END GPL.

From: SPSSX(r) Discussion [hidden email] On Behalf Of ViAnn Beadle
Sent: Friday, April 06, 2012 10:33 AM
To: [hidden email]
Subject: Re: Show only loess lines

You can also generate a chart showing only the loess lines using the GGRAPH command with GPL. Here’s an example assuming variables x and y:

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=x y MISSING=LISTWISE

REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: x=col(source(s), name("x"))

DATA: y=col(source(s), name("y"))

GUIDE: axis(dim(1), label("x label"))

GUIDE: axis(dim(2), label("y label"))

ELEMENT: line(position(smooth.loess(x*y)), missing.wings())

END GPL.

From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Art Kendall
Sent: Friday, April 06, 2012 9:15 AM
To: [hidden email]
Subject: Re: Show only loess lines

With the graph editor for the output window, you should be able to select the marker(s) for the points and either delete them of change them to no color at all.

Art Kendall

Social Research Consultants

On 4/6/2012 11:02 AM, William Adams wrote:

Hello Everyone,

Is it possible to show only smooth loess curves in a scatterplot? Basically, how can I hide or delete all the scatter dots and show only the loess lines?

Thank you!

Bill

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

William Adams

Re: Show only loess lines

Hi ViAnn,

Sorry – I am not very familiar with syntax. I tried this syntax below but am getting an error message that color.exterior(RIDRETH1) is unexpected.

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=RIDAGEYR MEAN(LBXSATSI)[name="MEAN_LBXSATSI"]

RIDRETH1 MISSING=LISTWISE REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: RIDAGEYR=col(source(s), name("RIDAGEYR"))

DATA: MEAN_LBXSATSI=col(source(s), name("MEAN_LBXSATSI"))

DATA: RIDRETH1=col(source(s), name("RIDRETH1"),

notIn("5"), unit.category())

GUIDE: axis(dim(1), label("Age"))

GUIDE: axis(dim(2), label("Mean ALT (U/L)"))

GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Race/Ethnicity - Recode"))

SCALE: cat(aesthetic(aesthetic.color.exterior), include("1", "2", "4", "6"))

ELEMENT: line(position(smooth.loess(RIDAGEYR*MEAN_LBXSATSI)), color.exterior(RIDRETH1)), missing.wings())

END GPL.

From: ViAnn Beadle [mailto:[hidden email]]
Sent: Friday, April 06, 2012 11:05 AM
To: [hidden email]
Cc: [hidden email]
Subject: RE: Show only loess lines

Sorry, I didn’t see any references to groups in your original post. Just add a color(RIDRETH1) function to your ELEMENT statement.

From: William Adams [[hidden email]]
Sent: Friday, April 06, 2012 9:54 AM
To: 'ViAnn Beadle'
Subject: RE: Show only loess lines

Hi ViAnn,

Thank you! Unfortunately, using the code below I am only seeing one line, but I need to see the loess lines for all of my groups. What am I missing?

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=RIDAGEYR MEAN(LBXSATSI)[name="MEAN_LBXSATSI"]

RIDRETH1 MISSING=LISTWISE REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: RIDAGEYR=col(source(s), name("RIDAGEYR"))

DATA: MEAN_LBXSATSI=col(source(s), name("MEAN_LBXSATSI"))

DATA: RIDRETH1=col(source(s), name("RIDRETH1"),

notIn("5"), unit.category())

GUIDE: axis(dim(1), label("Age"))

GUIDE: axis(dim(2), label("Mean ALT (U/L)"))

GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Race/Ethnicity - Recode"))

SCALE: cat(aesthetic(aesthetic.color.exterior), include("1", "2", "4", "6"))

ELEMENT: line(position(smooth.loess(RIDAGEYR*MEAN_LBXSATSI)), missing.wings())

END GPL.

From: SPSSX(r) Discussion [hidden email] On Behalf Of ViAnn Beadle
Sent: Friday, April 06, 2012 10:33 AM
To: [hidden email]
Subject: Re: Show only loess lines

You can also generate a chart showing only the loess lines using the GGRAPH command with GPL. Here’s an example assuming variables x and y:

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=x y MISSING=LISTWISE

REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: x=col(source(s), name("x"))

DATA: y=col(source(s), name("y"))

GUIDE: axis(dim(1), label("x label"))

GUIDE: axis(dim(2), label("y label"))

ELEMENT: line(position(smooth.loess(x*y)), missing.wings())

END GPL.

From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Art Kendall
Sent: Friday, April 06, 2012 9:15 AM
To: [hidden email]
Subject: Re: Show only loess lines

With the graph editor for the output window, you should be able to select the marker(s) for the points and either delete them of change them to no color at all.

Art Kendall

Social Research Consultants

On 4/6/2012 11:02 AM, William Adams wrote:

Hello Everyone,

Is it possible to show only smooth loess curves in a scatterplot? Basically, how can I hide or delete all the scatter dots and show only the loess lines?

Thank you!

Bill

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

David Marso

Re: Show only loess lines

Administrator

In reply to this post by William Adams

William!!!
***** PLEASE START A NEW TOPIC RATHER THAN JUMPING ON AN OLD ONE AND MODIFYING THE SUBJECT!!! THE THREADING IN NABBLE GETS COMPLETELY SCREWED UP!!!!!
************************************************************

William Adams wrote

Hello Everyone,

Is it possible to show only smooth loess curves in a scatterplot?
Basically, how can I hide or delete all the scatter dots and show only the
loess lines?

Thank you!

Bill

Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me.
---
"Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis."
Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?"

William Adams

Re: Show only loess lines

In reply to this post by ViAnn Beadle

Oh – I got it. Thanks ViAnn, and thank you everyone who contributed.

Have a great weekend!

Bill

From: William Adams [mailto:[hidden email]]
Sent: Friday, April 06, 2012 11:15 AM
To: 'ViAnn Beadle'
Cc: '[hidden email]'
Subject: RE: Show only loess lines

Hi ViAnn,

Sorry – I am not very familiar with syntax. I tried this syntax below but am getting an error message that color.exterior(RIDRETH1) is unexpected.

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=RIDAGEYR MEAN(LBXSATSI)[name="MEAN_LBXSATSI"]

RIDRETH1 MISSING=LISTWISE REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: RIDAGEYR=col(source(s), name("RIDAGEYR"))

DATA: MEAN_LBXSATSI=col(source(s), name("MEAN_LBXSATSI"))

DATA: RIDRETH1=col(source(s), name("RIDRETH1"),

notIn("5"), unit.category())

GUIDE: axis(dim(1), label("Age"))

GUIDE: axis(dim(2), label("Mean ALT (U/L)"))

GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Race/Ethnicity - Recode"))

SCALE: cat(aesthetic(aesthetic.color.exterior), include("1", "2", "4", "6"))

ELEMENT: line(position(smooth.loess(RIDAGEYR*MEAN_LBXSATSI)), color.exterior(RIDRETH1)), missing.wings())

END GPL.

From: ViAnn Beadle [[hidden email]]
Sent: Friday, April 06, 2012 11:05 AM
To: [hidden email]
Cc: [hidden email]
Subject: RE: Show only loess lines

Sorry, I didn’t see any references to groups in your original post. Just add a color(RIDRETH1) function to your ELEMENT statement.

From: William Adams [[hidden email]]
Sent: Friday, April 06, 2012 9:54 AM
To: 'ViAnn Beadle'
Subject: RE: Show only loess lines

Hi ViAnn,

Thank you! Unfortunately, using the code below I am only seeing one line, but I need to see the loess lines for all of my groups. What am I missing?

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=RIDAGEYR MEAN(LBXSATSI)[name="MEAN_LBXSATSI"]

RIDRETH1 MISSING=LISTWISE REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: RIDAGEYR=col(source(s), name("RIDAGEYR"))

DATA: MEAN_LBXSATSI=col(source(s), name("MEAN_LBXSATSI"))

DATA: RIDRETH1=col(source(s), name("RIDRETH1"),

notIn("5"), unit.category())

GUIDE: axis(dim(1), label("Age"))

GUIDE: axis(dim(2), label("Mean ALT (U/L)"))

GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Race/Ethnicity - Recode"))

SCALE: cat(aesthetic(aesthetic.color.exterior), include("1", "2", "4", "6"))

ELEMENT: line(position(smooth.loess(RIDAGEYR*MEAN_LBXSATSI)), missing.wings())

END GPL.

From: SPSSX(r) Discussion [hidden email] On Behalf Of ViAnn Beadle
Sent: Friday, April 06, 2012 10:33 AM
To: [hidden email]
Subject: Re: Show only loess lines

You can also generate a chart showing only the loess lines using the GGRAPH command with GPL. Here’s an example assuming variables x and y:

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=x y MISSING=LISTWISE

REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: x=col(source(s), name("x"))

DATA: y=col(source(s), name("y"))

GUIDE: axis(dim(1), label("x label"))

GUIDE: axis(dim(2), label("y label"))

ELEMENT: line(position(smooth.loess(x*y)), missing.wings())

END GPL.

From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Art Kendall
Sent: Friday, April 06, 2012 9:15 AM
To: [hidden email]
Subject: Re: Show only loess lines

With the graph editor for the output window, you should be able to select the marker(s) for the points and either delete them of change them to no color at all.

Art Kendall

Social Research Consultants

On 4/6/2012 11:02 AM, William Adams wrote:

Hello Everyone,

Is it possible to show only smooth loess curves in a scatterplot? Basically, how can I hide or delete all the scatter dots and show only the loess lines?

Thank you!

Bill

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

Ray Koopman

Re: Not sure whether nominal by interval test is appropriate

In reply to this post by Mike Donatello

----- Mike Donatello <[hidden email]> wrote:
> I should clarify that I do not have quality ratings for information
> sources which are not most-used. If I did, the problem would be
> simpler. (At least in my mind.)

And without those ratings there is nothing to correlate.

>
> MD
>
> On Thu, 5 Apr 2012 20:41:44 -0400, Mike Donatello
> <[hidden email]> wrote:
>
>> All,
>>
>> I'm having a difficult time with what should be a simple analysis.
>> I don't know whether I'm over-thinking the problem, but I can't
>> seem to figure out the appropriate method and syntax.
>>
>> I have (T=11) different topic areas, each of which has K different
>> potential sources of information. (The value of K varies among each
>> of the 11 topic areas.) That is, for each {T=1 . 11} topic(s) in
>> which a respondent is active there is the potential to have used
>> {1 . K} different information sources. Also, for the most frequently
>> used information source in each topic, the respondent provides a
>> rating of quality.
>>
>> What I am trying to find out -- within each single topic area in
>> which a respondent is active, and across all topic areas taken
>> together -- is whether there is a significant relationship between
>> the quality rating and an information source being reported as the
>> most frequently used.
>>
>> For example, say I am active in topics A, B and C. For topic A, I
>> used three sources; for topic B, I used nine sources; and for topic
>> C, I used two sources. I list my most-used source for A, B and C,
>> along with those sources' respective quality ratings. How do I know
>> whether the relationship between quality rating and a source being
>> listed as most-used is (1) significant for topic A, B or C, and
>> (2) significant regardless of topic?
>>
>> Initially, I was thinking that this was a simple case of nominal
>> (binary score for most-used or not) by interval (quality rating)
>> correlation, but now I'm not sure whether that's the proper approach.
>> I have a nagging feeling that I am discarding useful information by
>> going in that direction. (My dataset is structured in a way that will
>> make that test a bear, but I think I can pull it off.) I assume that
>> either logistic regression or point biserial correlation would be
>> appropriate in that case.
>>
>> Is there some other method that I am overlooking?
>>
>> Thanks in advance for any help.
>>
>> ---
>>
>> Mike Donatello
>>
>> 703.582.5680
>> <mailto:[hidden email]> [hidden email]

Mike Donatello

Re: Not sure whether nominal by interval test is appropriate

Yup, I realize that. I decided to test instead whether the mean rating for most-used sites is higher than the midpoint of the scale. Not the same question, but close enough for my needs.

Thanks.

-----Original Message-----
From: Ray Koopman [mailto:[hidden email]]
Sent: Friday, April 06, 2012 12:35 PM
To: Mike Donatello
Cc: [hidden email]
Subject: Re: Not sure whether nominal by interval test is appropriate

----- Mike Donatello <[hidden email]> wrote:
> I should clarify that I do not have quality ratings for information
> sources which are not most-used. If I did, the problem would be
> simpler. (At least in my mind.)

And without those ratings there is nothing to correlate.

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