Scanning grouped data for outliers

[Benjamin Spivak (Med)]

Hi everybody,

I assume there is a very simple solution to this problem. I simply cannot find a way to scan for outliers in groups. Can somebody enlighten me?

Regards,

Ben.

[David Marso]

The road to enlightenment is paved with such inconveniences as taking the time to describe your issue with some specificity?  Perhaps try SPLIT FILE and EXAMINE?

Benjamin Spivak (Med) wrote

Hi everybody,

I assume there is a very simple solution to this problem. I simply cannot
find a way to scan for outliers in groups. Can somebody enlighten me?

Regards,

Ben.

[Benjamin Spivak (Med)]

Haha, my apologies, David. 

I want to screen my data for a 2x3 experiment using multiple binary and ordinal DV's (Approx 15 in total). before I engage in analysis proper, I wanted to screen my data for univariate outliers, multivariate outliers, normality, heterogeneity of variance etc. 

My data is nested ( 63 groups with 10-12 subjects each), I don't know if that presents a complication for screening.

If there are any further details required please let me know.

Regards,

Ben.

On 27 May 2012 16:47, David Marso <[hidden email]> wrote:

The road to enlightenment is paved with such inconveniences as taking the
time to describe your issue with some specificity?  Perhaps try SPLIT FILE
and EXAMINE?

Benjamin Spivak (Med) wrote

>
> Hi everybody,
>
> I assume there is a very simple solution to this problem. I simply cannot
> find a way to scan for outliers in groups. Can somebody enlighten me?
>
> Regards,
>
> Ben.
>

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Scanning-grouped-data-for-outliers-tp5713388p5713389.html
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

[Jon K Peck]

There are several tools in Statistics that might help here.  For example,
EXAMINE reports outliers and can process groups.
Anomaly detection, aka Identify Unusual Cases uses clustering techniques for this.  It doesn't do groups explicitly, but you can include the grouping variables in the analysis variables list.

Graphically, you might look at paneled plots, including the STATS SUBGROUP PLOTS extension command available from the SPSS Community website.

And, of course, residuals analysis from various procedures would be useful.

HTH,

Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
new phone: [phone]




From:        Benjamin Spivak <[hidden email]>
To:        [hidden email]
Date:        05/26/2012 11:44 PM
Subject:        [SPSSX-L] Scanning grouped data for outliers
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

Hi everybody,

I assume there is a very simple solution to this problem. I simply cannot find a way to scan for outliers in groups. Can somebody enlighten me?

Regards,

Ben.

[news]

Hi Ben,
What about James Schwab's methods page at
http://www.utexas.edu/courses/schwab/sw388r7/SolvingProblems/SolvingHomeworkProblems.htm

and in particular
http://www.utexas.edu/courses/schwab/sw388r7/SolvingProblems/DetectingOutliers.ppt
?

HTH

F. Thomas

On 27/05/2012 07:39, Benjamin Spivak wrote:
> Hi everybody,
>
> I assume there is a very simple solution to this problem. I simply
> cannot find a way to scan for outliers in groups. Can somebody
> enlighten me?
>
> Regards,
>
> Ben.

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

[Benjamin Spivak (Med)]

Thanks everybody. I will try these methods tonight and reply.

On 28 May 2012 17:52, ftr <[hidden email]> wrote:

Hi Ben,
What about James Schwab's methods page at
http://www.utexas.edu/courses/schwab/sw388r7/SolvingProblems/SolvingHomeworkProblems.htm

and in particular
http://www.utexas.edu/courses/schwab/sw388r7/SolvingProblems/DetectingOutliers.ppt
?

HTH

F. Thomas

On 27/05/2012 07:39, Benjamin Spivak wrote:

Hi everybody,

I assume there is a very simple solution to this problem. I simply
cannot find a way to scan for outliers in groups. Can somebody
enlighten me?

Regards,

Ben.

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

The very first step in working on suspicious  (anomalous) values is to proofread or double enter the data.
YMMV but in my experience a subjective 80-some percent are data entry errors.

On 5/27/2012 1:39 AM, Benjamin Spivak wrote:

Hi everybody,

I assume there is a very simple solution to this problem. I simply cannot find a way to scan for outliers in groups. Can somebody enlighten me?

Regards,

Ben.

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

[zstatman]

There are several ways to do this like double entry and compare, as Art noted, to using statistical measures such as Cook's in SPSS so exactly at what level are you seeking to do this?

 

WMB
Statistical Services

============
mailto: [hidden email]
http:\\home.earthlink.net\~info.statman
Skype & Google Talk: zstatman
============

---

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Art Kendall
Sent: Monday, May 28, 2012 7:10 AM
To: [hidden email]
Subject: Re: Scanning grouped data for outliers

The very first step in working on suspicious  (anomalous) values is to proofread or double enter the data.
YMMV but in my experience a subjective 80-some percent are data entry errors.

On 5/27/2012 1:39 AM, Benjamin Spivak wrote:

Hi everybody,

I assume there is a very simple solution to this problem. I simply cannot find a way to scan for outliers in groups. Can somebody enlighten me?

Regards,

Ben.

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