How to set a caliper width for propensity score matching

Dear experts,

We are analyzing data using propensity matching. We would like to set a
caliper width at 0.2. In the propensity matching menu, there is a section
named 'tolerance'. Is this the same as a caliper width? If not, is it
possible to convert a value for a caliper width to one for tolerance?
Thank you for your help.
Regards,
Hideo



--
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

[William Dudley-2]

Hideo,

I have not used, and cannot comment on, the SPSS propensity score method.

However, you may find the R Program MatchIt useful because it provides several methods in addition to propensity score matching.

https://gking.harvard.edu/matchit

The down sides include; a learning curve with using R, and the fact that you have to deal with some file management issues getting data from SPSS and into R.

But, it might be that the wider array of matching algorithms in MatchIt, and good documentation make it worth looking at.

Bill

.

On Sun, Apr 15, 2018 at 11:07 PM, [hidden email] <[hidden email]> wrote:

Dear experts,

We are analyzing data using propensity matching. We would like to set a
caliper width at 0.2. In the propensity matching menu, there is a section
named 'tolerance'. Is this the same as a caliper width? If not, is it
possible to convert a value for a caliper width to one for tolerance?
Thank you for your help.
Regards,
Hideo



--
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

--

===================== 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]

Hideo,

SPSS Tolerance and Calipers are not the same. Calipers is based on the standard deviation of the logit of the propensity score, so a caliper of 0.20 for example, means allowing for a difference of .20 standard deviations. Tolerance is expressed as a proportion of the propensity score, so a tolerance of 0.20 means allowing for a difference of .20 in the overall propensity score.

Felix Thoemmes has a paper at the link below, which describes using a package in R with the SPSS R plug-in, which will allow you to use calipers.

https://arxiv.org/ftp/arxiv/papers/1201/1201.6385.pdf

HTH.

Brian

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of William Dudley <[hidden email]>
Sent: Monday, April 16, 2018 9:37:54 AM
To: [hidden email]
Subject: Re: How to set a caliper width for propensity score matching

 

Hideo,

I have not used, and cannot comment on, the SPSS propensity score method.

However, you may find the R Program MatchIt useful because it provides several methods in addition to propensity score matching.

https://gking.harvard.edu/matchit

The down sides include; a learning curve with using R, and the fact that you have to deal with some file management issues getting data from SPSS and into R.

But, it might be that the wider array of matching algorithms in MatchIt, and good documentation make it worth looking at.

Bill

.

On Sun, Apr 15, 2018 at 11:07 PM, [hidden email] <[hidden email]> wrote:

Dear experts,

We are analyzing data using propensity matching. We would like to set a
caliper width at 0.2. In the propensity matching menu, there is a section
named 'tolerance'. Is this the same as a caliper width? If not, is it
possible to convert a value for a caliper width to one for tolerance?
Thank you for your help.
Regards,
Hideo



--
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

--

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

[Jon Peck]

The SPSS matching procedure dialog has a field labelled Match Tolerance, so I presume that it is the SPSS procedure you are referring to.  Match Tolerance is not the same as Caliper in the R-based procedure.

The R-based procedure has a dialog for PS Matching on the Analyze menu once you install that as William says.  But once installed, it will act like an SPSS procedure, and there should be no R knowledge or manipulations required in order to use it.  The dialog does not have a help button, however.

On Mon, Apr 16, 2018 at 7:37 AM, William Dudley <[hidden email]> wrote:

Hideo,

I have not used, and cannot comment on, the SPSS propensity score method.

However, you may find the R Program MatchIt useful because it provides several methods in addition to propensity score matching.

https://gking.harvard.edu/matchit

The down sides include; a learning curve with using R, and the fact that you have to deal with some file management issues getting data from SPSS and into R.

But, it might be that the wider array of matching algorithms in MatchIt, and good documentation make it worth looking at.

Bill

.

On Sun, Apr 15, 2018 at 11:07 PM, [hidden email] <[hidden email]> wrote:

Dear experts,

We are analyzing data using propensity matching. We would like to set a
caliper width at 0.2. In the propensity matching menu, there is a section
named 'tolerance'. Is this the same as a caliper width? If not, is it
possible to convert a value for a caliper width to one for tolerance?
Thank you for your help.
Regards,
Hideo



--
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

--

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

Thank you very much for your comments.

I understood the tolerance was different from a caliper width. I think we
are going to use R-based one. Thank you again for your inputs.

Regards,
Hideo



--
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