Correlation between ordinal data
Locked
 
|
I have two columns of data made up of 10 rows, representing families. The data in each row is the choice (Frequently (3), Moderately (2), and Rarely (1)) made by a family member and professional about another individual. I am interested if there is a correlation between the two: family members and professionals.
Does a Spearman rho provide me with this answer or is there a better test? TIA, Stephen Salbod, Pace University, NYC |
Re: Correlation between ordinal data
|
|
Maybe you can find useful the polychoric correlation:
Regards!! 2014-04-23 10:02 GMT-05:00 Salbod <[hidden email]>: I have two columns of data made up of 10 rows, representing families. The |
|
|
Thanks Norberto. I thought about the polychoric, but I couldn’t find it in SPSS and I wasn’t sure if it used the ordering information. I’ll check out the link
you so kindly provided. --Steve From: SPSSX(r) Discussion [mailto:[hidden email]]
On Behalf Of Norberto Hernandez Maybe you can find useful the polychoric correlation: Regards!! 2014-04-23 10:02 GMT-05:00 Salbod <[hidden email]>:
|
|
|
In reply to this post by Salbod
Check out the HETCOR extension bundle on the IBM / SPSS website.
This procedure calculates correlations between nominal, ordinal, and scale variables, accounting for the measurement levels of the variables. The resulting heterogeneous correlation matrix consists of Pearson product-moment correlations between scale variables, polyserial correlations between scale and categorical variables, and polychoric correlations between categorical variables. The procedure uses the hetcor function from the R polycor package. On Wed, 23 Apr 2014 08:02:42 -0700, Salbod <[hidden email]> wrote: >I have two columns of data made up of 10 rows, representing families. The >data in each row is the choice (Frequently (3), Moderately (2), and Rarely >(1)) made by a family member and professional about another individual. I >am interested if there is a correlation between the two: family members and >professionals. >Does a Spearman rho provide me with this answer or is there a better test? > >TIA, Stephen Salbod, Pace University, NYC > > > > >-- >View this message in context: >Sent from the SPSSX Discussion mailing list archive at 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 ===================== 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 |
Re: Correlation between ordinal data
|
|
In reply to this post by Salbod
Polychoric correlation is available in
the SPSSINC HETCOR extension command - it computes Pearson, polyserial,
or polychoric depending on the measurement levels of the variables.
Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] phone: 720-342-5621 From: "Salbod, Mr. Stephen" <[hidden email]> To: [hidden email], Date: 04/23/2014 09:41 AM Subject: Re: [SPSSX-L] Correlation between ordinal data Sent by: "SPSSX(r) Discussion" <[hidden email]> Thanks Norberto. I thought about the polychoric, but I couldn’t find it in SPSS and I wasn’t sure if it used the ordering information. I’ll check out the link you so kindly provided. --Steve From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Norberto Hernandez Sent: Wednesday, April 23, 2014 11:11 AM To: [hidden email] Subject: Re: Correlation between ordinal data Maybe you can find useful the polychoric correlation: http://john-uebersax.com/stat/tetra.htm Regards!! 2014-04-23 10:02 GMT-05:00 Salbod <ssalbod@...>: I have two columns of data made up of 10 rows, representing families. The data in each row is the choice (Frequently (3), Moderately (2), and Rarely (1)) made by a family member and professional about another individual. I am interested if there is a correlation between the two: family members and professionals. Does a Spearman rho provide me with this answer or is there a better test? TIA, Stephen Salbod, Pace University, NYC -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Correlation-between-ordinal-data-tp5725627.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSX-L, send a message to LISTSERV@... (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 |
Re: Correlation between ordinal data
|
|
In reply to this post by Norberto Hernandez
I'm pretty sure that I would stick with the plain, old Pearson r.
Spearman is much better suited to continuous data: Is there really any improvement in anything by transforming (1,2,3) into the "average rank" observed with massive numbers of ties? Of course, if you ask for a Monte Carlo evaluation, you do get an "exact" test. Otherwise, the tests on Spearman are only as good as the adjustment-in-error made for ties (which is not always so good). My impression of the tetrachoric and polychoric correlations is that they have a role where theorists are paying close attention to using r as a precise measure of the "underlying relationships". The price of this gain is that you don't have any decent tests... and you probably mislead your readership, who fail to note that these numbers are inflated, compared to the r's that they are familiar with for similar situations. -- Rich Ulrich Date: Wed, 23 Apr 2014 10:10:32 -0500 From: [hidden email] Subject: Re: Correlation between ordinal data To: [hidden email] Maybe you can find useful the polychoric correlation: Regards!! 2014-04-23 10:02 GMT-05:00 Salbod <[hidden email]>: I have two columns of data made up of 10 rows, representing families. The |
|
|
That seems like a fair assessment to me. I've seen arguments for using Kendall's Tau over Spearman when comparing treatments, see this discussion and other linked questions on Crossvalidated, Kendall Tau or Spearman's rho?. But for just estimating the correlation between the two measures any one of them may be sufficient.
With only three measures for each a plot (and/or a table) is fine as well in my opinion. Especially with only 10 observations. |
Re: Correlation between ordinal data
|
In reply to this post by Salbod
With such a coarse
measurement scale and only 10 cases, I would take any result
with a good dose of salt even with the exact tests.
Art Kendall Social Research ConsultantsOn 4/23/2014 11:02 AM, Salbod [via SPSSX Discussion] wrote: I have two columns of data made up of 10 rows, representing families. The data in each row is the choice (Frequently (3), Moderately (2), and Rarely (1)) made by a family member and professional about another individual. I am interested if there is a correlation between the two: family members and professionals.
Art Kendall
Social Research Consultants |
|
|
In reply to this post by Rich Ulrich
Thanks Rich. I went with the Pearson r—as you said it’s plain. -Steve From: SPSSX(r) Discussion [mailto:[hidden email]]
On Behalf Of Rich Ulrich I'm pretty sure that I would stick with the plain, old Pearson r.
Date: Wed, 23 Apr 2014 10:10:32 -0500 Maybe you can find useful the polychoric correlation: Regards!! 2014-04-23 10:02 GMT-05:00 Salbod <[hidden email]>:
|
|
|
In reply to this post by Jon K Peck
Jon, Thanks for mentioning SPSSINC HETCOR. Are there instructions on how I can access that procedure? Thank you. -Steve From: Jon K Peck [mailto:[hidden email]]
Polychoric correlation is available in the SPSSINC HETCOR extension command - it computes Pearson, polyserial, or polychoric depending on the measurement levels of the variables.
|
|
|
In reply to this post by Art Kendall
Art: Thank you for the smile
J Steve From: SPSSX(r) Discussion [mailto:[hidden email]]
On Behalf Of Art Kendall With such a coarse measurement scale and only 10 cases, I would take any result with a good dose of salt even with the exact tests. Art Kendall Social Research Consultants On 4/23/2014 11:02 AM, Salbod [via SPSSX Discussion] wrote:
Art Kendall View this message in context:
Re: Correlation between ordinal data |
| Free forum by Nabble | Edit this page |