SPSSX Discussion

Not Enough Cases To Perform Cluster Analysis

Classic

List

Threaded

7 messages Options

Colin Valdiserri

Not Enough Cases To Perform Cluster Analysis

Hello everyone,

As you can tell from the subject line, I am running some cluster analyses using K-Means in SPSS 18 (PASW) and I am getting and error when I run the analyses that tells me “Not Enough Cases To Perform Cluster Analysis” – “Execution of this command stops”. The problem is that I do have plenty of cases; a total weighted sample of N=3,268 to be exact. Now, some of the variables do have missing values because of the skip pattern in the survey. But there are still plenty of cases, or so I would think.

Out of the 12 variables in one of the cluster analyses I have the following number of cases for each variable:

· Q21_1 à Number of valid cases is 672

· Q21_2 à Number of valid cases is 746

· Q21_3 à Number of valid cases is 1833

· Q21_4 à Number of valid cases is 2265

· Q21_5 à Number of valid cases is 2414

· Q21_6 à Number of valid cases is 2597

· Q21_7 à Number of valid cases is 2704

· Q21_8 à Number of valid cases is 1413

· Q21_9 à Number of valid cases is 1423

· Q21_10 à Number of valid cases is 802

· Q21_11 à Number of valid cases is 2404

· Q21_12 à Number of valid cases is 1610

My syntax is below:

QUICK CLUSTER q21_1 q21_2 q21_3 q21_4 q21_5 q21_6 q21_7 q21_8 q21_9 q21_10

q21_11 q21_12

/MISSING=PAIRWISE

/CRITERIA=CLUSTER(5) MXITER(100) CONVERGE(0)

/METHOD=KMEANS(NOUPDATE)

/PRINT INITIAL.

I don’t understand why it is doing this or if there is anything I can do to get it to run the cluster analysis. I would love to hear any thoughts.

Thank you in advance,

Colin

Robert Walker

Re: Not Enough Cases To Perform Cluster Analysis

Colin,

We need to know more – what percent of sample have data for all 12 variables? Conversely, are there missing values for every respondent? On just one attribute? On multiple attributes? What was the skip logic? I am assuming that Q21_1 to Q21_12 are attributes using some sort of agreement rating scale? What was the scale and scale values?

To trigger that error, you must have a missing value for every respondent on at least one variable. If so, you can't run this procedure without recoding that missing data, either through some form of substitution to populate those missing values, or a RECODE statement to assign them a usable value.

Regards,

Bob Walker

Surveys & Forecasts, LLC

www.safllc.com

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Colin Valdiserri
Sent: Saturday, September 21, 2013 2:03 PM
To: [hidden email]
Subject: Not Enough Cases To Perform Cluster Analysis

Hello everyone,

Out of the 12 variables in one of the cluster analyses I have the following number of cases for each variable:

· Q21_1 à Number of valid cases is 672

· Q21_2 à Number of valid cases is 746

· Q21_3 à Number of valid cases is 1833

· Q21_4 à Number of valid cases is 2265

· Q21_5 à Number of valid cases is 2414

· Q21_6 à Number of valid cases is 2597

· Q21_7 à Number of valid cases is 2704

· Q21_8 à Number of valid cases is 1413

· Q21_9 à Number of valid cases is 1423

· Q21_10 à Number of valid cases is 802

· Q21_11 à Number of valid cases is 2404

· Q21_12 à Number of valid cases is 1610

My syntax is below:

QUICK CLUSTER q21_1 q21_2 q21_3 q21_4 q21_5 q21_6 q21_7 q21_8 q21_9 q21_10

q21_11 q21_12

/MISSING=PAIRWISE

/CRITERIA=CLUSTER(5) MXITER(100) CONVERGE(0)

/METHOD=KMEANS(NOUPDATE)

/PRINT INITIAL.

I don’t understand why it is doing this or if there is anything I can do to get it to run the cluster analysis. I would love to hear any thoughts.

Thank you in advance,

Colin

Colin Valdiserri

Re: Not Enough Cases To Perform Cluster Analysis

You answered my question right there… I do have a few cases who have missing values for all of those variables and didn’t realize. Thank you!

Colin M. Valdiserri

Co-Founder & CEO

Informed Decisions Group, Inc.

8854 Jordan Court | North Ridgeville, Ohio | 44039
P: 440.935.5414
F: 440.353.0621

cvaldiserri@...
www.idg-consulting.com

The information and attachments contained in this e-mail message is intended for the use of the recipient(s) named above and is privileged and confidential. If you are not the intended recipient, you are formally notified that you have received this message in error and that any review, dissemination, distribution, or copying of the message is strictly prohibited. If you have received this communication in error, please notify us immediately by e-mail and delete the original message.

From: Bob Walker [mailto:[hidden email]]
Sent: Saturday, September 21, 2013 2:30 PM
To: Colin Valdiserri; [hidden email]
Subject: RE: Not Enough Cases To Perform Cluster Analysis

Colin,

Regards,

Bob Walker

Surveys & Forecasts, LLC

www.safllc.com

From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Colin Valdiserri
Sent: Saturday, September 21, 2013 2:03 PM
To: [hidden email]
Subject: Not Enough Cases To Perform Cluster Analysis

Hello everyone,

Out of the 12 variables in one of the cluster analyses I have the following number of cases for each variable:

· Q21_1 à Number of valid cases is 672

· Q21_2 à Number of valid cases is 746

· Q21_3 à Number of valid cases is 1833

· Q21_4 à Number of valid cases is 2265

· Q21_5 à Number of valid cases is 2414

· Q21_6 à Number of valid cases is 2597

· Q21_7 à Number of valid cases is 2704

· Q21_8 à Number of valid cases is 1413

· Q21_9 à Number of valid cases is 1423

· Q21_10 à Number of valid cases is 802

· Q21_11 à Number of valid cases is 2404

· Q21_12 à Number of valid cases is 1610

My syntax is below:

QUICK CLUSTER q21_1 q21_2 q21_3 q21_4 q21_5 q21_6 q21_7 q21_8 q21_9 q21_10

q21_11 q21_12

/MISSING=PAIRWISE

/CRITERIA=CLUSTER(5) MXITER(100) CONVERGE(0)

/METHOD=KMEANS(NOUPDATE)

/PRINT INITIAL.

I don’t understand why it is doing this or if there is anything I can do to get it to run the cluster analysis. I would love to hear any thoughts.

Thank you in advance,

Colin

Jon K Peck

Re: Not Enough Cases To Perform Cluster Analysis

The MVA procedure in the Missing Values option (Analyze > Missing Value Analysis) can be very helpful for identifying patterns of missingness.

Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
phone: 720-342-5621

From: Colin Valdiserri <[hidden email]>
To: [hidden email],
Date: 09/21/2013 12:38 PM
Subject: Re: [SPSSX-L] Not Enough Cases To Perform Cluster Analysis
Sent by: "SPSSX(r) Discussion" <[hidden email]>

You answered my question right there… I do have a few cases who have missing values for all of those variables and didn’t realize. Thank you!

Colin M. Valdiserri
Co-Founder & CEO
Informed Decisions Group, Inc.
8854 Jordan Court | North Ridgeville, Ohio | 44039
P: 440.935.5414
F: 440.353.0621
cvaldiserri@...
www.idg-consulting.com

Follow us on LinkedIn
The information and attachments contained in this e-mail message is intended for the use of the recipient(s) named above and is privileged and confidential. If you are not the intended recipient, you are formally notified that you have received this message in error and that any review, dissemination, distribution, or copying of the message is strictly prohibited. If you have received this communication in error, please notify us immediately by e-mail and delete the original message.

From: Bob Walker [mailto:rww@...]
Sent: Saturday, September 21, 2013 2:30 PM
To: Colin Valdiserri; [hidden email]
Subject: RE: Not Enough Cases To Perform Cluster Analysis

Colin,

We need to know more – what percent of sample have data for all 12 variables? Conversely, are there missing values for every respondent? On just one attribute? On multiple attributes? What was the skip logic? I am assuming that Q21_1 to Q21_12 are attributes using some sort of agreement rating scale? What was the scale and scale values?

To trigger that error, you must have a missing value for every respondent on at least one variable. If so, you can't run this procedure without recoding that missing data, either through some form of substitution to populate those missing values, or a RECODE statement to assign them a usable value.

Regards,

Bob Walker
Surveys & Forecasts, LLC
www.safllc.com

From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Colin Valdiserri
Sent: Saturday, September 21, 2013 2:03 PM
To: [hidden email]
Subject: Not Enough Cases To Perform Cluster Analysis

Hello everyone,

As you can tell from the subject line, I am running some cluster analyses using K-Means in SPSS 18 (PASW) and I am getting and error when I run the analyses that tells me “Not Enough Cases To Perform Cluster Analysis” – “Execution of this command stops”. The problem is that I do have plenty of cases; a total weighted sample of N=3,268 to be exact. Now, some of the variables do have missing values because of the skip pattern in the survey. But there are still plenty of cases, or so I would think.

Out of the 12 variables in one of the cluster analyses I have the following number of cases for each variable:
· Q21_1 à Number of valid cases is 672
· Q21_2 à Number of valid cases is 746
· Q21_3 à Number of valid cases is 1833
· Q21_4 à Number of valid cases is 2265
· Q21_5 à Number of valid cases is 2414
· Q21_6 à Number of valid cases is 2597
· Q21_7 à Number of valid cases is 2704
· Q21_8 à Number of valid cases is 1413
· Q21_9 à Number of valid cases is 1423
· Q21_10 à Number of valid cases is 802
· Q21_11 à Number of valid cases is 2404
· Q21_12 à Number of valid cases is 1610

My syntax is below:
QUICK CLUSTER q21_1 q21_2 q21_3 q21_4 q21_5 q21_6 q21_7 q21_8 q21_9 q21_10
q21_11 q21_12
/MISSING=PAIRWISE
/CRITERIA=CLUSTER(5) MXITER(100) CONVERGE(0)
/METHOD=KMEANS(NOUPDATE)
/PRINT INITIAL.

I don’t understand why it is doing this or if there is anything I can do to get it to run the cluster analysis. I would love to hear any thoughts.

Thank you in advance,

Colin

John F Hall

Re: Not Enough Cases To Perform Cluster Analysis

In reply to this post by Colin Valdiserri

Colin

It might help to check the pattern of missing responses. To find the total of missing values within a case across the list of q21_1 to q21_12.

(You need to change the smart quotes: SPSS doesn’t like them)

count xmis = q21_1 to q21_12 (sysmis)

/ymis = q21_1 to q21_12 (missing).

format xmis ymis (f4.0).

var lab xmis ‘Number of cases with system missing)’

/ymis ‘Number of cases with both system and user missing’.

freq xmis ymis.

cro q21_1 to q21_12 by xmis ymis.

You can do cluster analysis manually using the correlation matrix:

corr q21_1 to q21_12 .

1: Find the highest r in the matrix and circle it.

2: Look along the row with the circled r and draw a circle round any r that is the highest with that var for any other variable.

3: For each of these look up and down the respective column(s) to see if any unchecked variable has its highest r with one of those. Circle those.

4: Look along the rows of these and repeat step one above.

You can do this starting in a column or a row as long as you alternate the direction of the search at each step. If you reach a point when no other variable has its highest r with one of the ones already circled, you’ve reached the end of your first cluster.

Repeat for any remaining variables starting with the highest remaing r, then check to see if the clusters make sense.

I know a disgusting joke to explain the underlying idea. It involves a spittoon, but vividly illustrates the nature of a cluster: my students loved it.

John F Hall (Mr)

[Retired academic survey researcher]

Email: [hidden email]

Website: www.surveyresearch.weebly.com

SPSS start page: www.surveyresearch.weebly.com/spss-without-tears.html

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Colin Valdiserri
Sent: 21 September 2013 20:37
To: [hidden email]
Subject: Re: Not Enough Cases To Perform Cluster Analysis

You answered my question right there… I do have a few cases who have missing values for all of those variables and didn’t realize. Thank you!

Colin M. Valdiserri

Co-Founder & CEO

Informed Decisions Group, Inc.

8854 Jordan Court | North Ridgeville, Ohio | 44039
P: 440.935.5414
F: 440.353.0621

cvaldiserri@...
www.idg-consulting.com

From: Bob Walker [[hidden email]]
Sent: Saturday, September 21, 2013 2:30 PM
To: Colin Valdiserri; [hidden email]
Subject: RE: Not Enough Cases To Perform Cluster Analysis

Colin,

Regards,

Bob Walker

Surveys & Forecasts, LLC

www.safllc.com

Hello everyone,

Out of the 12 variables in one of the cluster analyses I have the following number of cases for each variable:

· Q21_1 à Number of valid cases is 672

· Q21_2 à Number of valid cases is 746

· Q21_3 à Number of valid cases is 1833

· Q21_4 à Number of valid cases is 2265

· Q21_5 à Number of valid cases is 2414

· Q21_6 à Number of valid cases is 2597

· Q21_7 à Number of valid cases is 2704

· Q21_8 à Number of valid cases is 1413

· Q21_9 à Number of valid cases is 1423

· Q21_10 à Number of valid cases is 802

· Q21_11 à Number of valid cases is 2404

· Q21_12 à Number of valid cases is 1610

My syntax is below:

QUICK CLUSTER q21_1 q21_2 q21_3 q21_4 q21_5 q21_6 q21_7 q21_8 q21_9 q21_10

q21_11 q21_12

/MISSING=PAIRWISE

/CRITERIA=CLUSTER(5) MXITER(100) CONVERGE(0)

/METHOD=KMEANS(NOUPDATE)

/PRINT INITIAL.

I don’t understand why it is doing this or if there is anything I can do to get it to run the cluster analysis. I would love to hear any thoughts.

Thank you in advance,

Colin

Art Kendall

Re: Not Enough Cases To Perform Cluster Analysis

That was one of the McQuitty methods.

I had an old FORTRAN program in the 70's that did it.

I have it on my "things to do" someday list. It is there with his "iterative correlation analysis" where you correlate cases and then correlate rows iteratively.

Art Kendall
Social Research Consultants

On 9/21/2013 11:11 PM, John F Hall [via SPSSX Discussion] wrote:

Colin

It might help to check the pattern of missing responses. To find the total of missing values within a case across the list of q21_1 to q21_12.

(You need to change the smart quotes: SPSS doesn’t like them)

count xmis = q21_1 to q21_12 (sysmis)

          /ymis = q21_1 to q21_12 (missing).

format xmis ymis (f4.0).

var lab xmis ‘Number of cases with system missing)’

            /ymis ‘Number of cases with both system and user missing’.

freq xmis ymis.

cro q21_1 to q21_12 by xmis ymis.

You can do cluster analysis manually using the correlation matrix:

corr q21_1 to q21_12 .

1: Find the highest r in the matrix and circle it.

2: Look along the row with the circled r and draw a circle round any r that is the highest with that var for any other variable.

3: For each of these look up and down the respective column(s) to see if any unchecked variable has its highest r with one of those. Circle those.

4: Look along the rows of these and repeat step one above.

You can do this starting in a column or a row as long as you alternate the direction of the search at each step. If you reach a point when no other variable has its highest r with one of the ones already circled, you’ve reached the end of your first cluster.

Repeat for any remaining variables starting with the highest remaing r, then check to see if the clusters make sense.

I know a disgusting joke to explain the underlying idea. It involves a spittoon, but vividly illustrates the nature of a cluster: my students loved it.

John F Hall (Mr)

[Retired academic survey researcher]

Email:   [hidden email]

Website: www.surveyresearch.weebly.com

SPSS start page: www.surveyresearch.weebly.com/spss-without-tears.html





From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Colin Valdiserri
Sent: 21 September 2013 20:37
To: [hidden email]
Subject: Re: Not Enough Cases To Perform Cluster Analysis

You answered my question right there… I do have a few cases who have missing values for all of those variables and didn’t realize. Thank you!

Colin M. Valdiserri

Co-Founder & CEO

Informed Decisions Group, Inc.

8854 Jordan Court | North Ridgeville, Ohio | 44039
P: 440.935.5414
F: 440.353.0621

cvaldiserri@...
www.idg-consulting.com

Follow us on LinkedIn

The information and attachments contained in this e-mail message is intended for the use of the recipient(s) named above and is privileged and confidential. If you are not the intended recipient, you are formally notified that you have received this message in error and that any review, dissemination, distribution, or copying of the message is strictly prohibited. If you have received this communication in error, please notify us immediately by e-mail and delete the original message.

From: Bob Walker [[hidden email]]
Sent: Saturday, September 21, 2013 2:30 PM
To: Colin Valdiserri; [hidden email]
Subject: RE: Not Enough Cases To Perform Cluster Analysis

Colin,

We need to know more – what percent of sample have data for all 12 variables? Conversely, are there missing values for every respondent? On just one attribute? On multiple attributes? What was the skip logic? I am assuming that Q21_1 to Q21_12 are attributes using some sort of agreement rating scale? What was the scale and scale values?

To trigger that error, you must have a missing value for every respondent on at least one variable. If so, you can't run this procedure without recoding that missing data, either through some form of substitution to populate those missing values, or a RECODE statement to assign them a usable value.

Regards,

Bob Walker

Surveys & Forecasts, LLC

www.safllc.com

From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Colin Valdiserri
Sent: Saturday, September 21, 2013 2:03 PM
To: [hidden email]
Subject: Not Enough Cases To Perform Cluster Analysis

Hello everyone,

As you can tell from the subject line, I am running some cluster analyses using K-Means in SPSS 18 (PASW) and I am getting and error when I run the analyses that tells me “Not Enough Cases To Perform Cluster Analysis” – “Execution of this command stops”. The problem is that I do have plenty of cases; a total weighted sample of N=3,268 to be exact. Now, some of the variables do have missing values because of the skip pattern in the survey. But there are still plenty of cases, or so I would think.

Out of the 12 variables in one of the cluster analyses I have the following number of cases for each variable:

·         Q21_1 à Number of valid cases is 672

·         Q21_2 à Number of valid cases is 746

·         Q21_3 à Number of valid cases is 1833

·         Q21_4 à Number of valid cases is 2265

·         Q21_5 à Number of valid cases is 2414

·         Q21_6 à Number of valid cases is 2597

·         Q21_7 à Number of valid cases is 2704

·         Q21_8 à Number of valid cases is 1413

·         Q21_9 à Number of valid cases is 1423

·         Q21_10 à Number of valid cases is 802

·         Q21_11 à Number of valid cases is 2404

·         Q21_12 à Number of valid cases is 1610

My syntax is below:

QUICK CLUSTER q21_1 q21_2 q21_3 q21_4 q21_5 q21_6 q21_7 q21_8 q21_9 q21_10

    q21_11 q21_12

  /MISSING=PAIRWISE

/CRITERIA=CLUSTER(5) MXITER(100) CONVERGE(0)

/METHOD=KMEANS(NOUPDATE)

/PRINT INITIAL.

I don’t understand why it is doing this or if there is anything I can do to get it to run the cluster analysis. I would love to hear any thoughts.

Thank you in advance,

Colin

If you reply to this email, your message will be added to the discussion below:
http://spssx-discussion.1045642.n5.nabble.com/Not-Enough-Cases-To-Perform-Cluster-Analysis-tp5722164p5722169.html

To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

Art Kendall
Social Research Consultants

Art Kendall

Re: Not Enough Cases To Perform Cluster Analysis

In reply to this post by Colin Valdiserri

Weights are not taken into account in clustering.

Art Kendall
Social Research Consultants

On 9/21/2013 2:04 PM, Colin Valdiserri [via SPSSX Discussion] wrote:

Hello everyone,

As you can tell from the subject line, I am running some cluster analyses using K-Means in SPSS 18 (PASW) and I am getting and error when I run the analyses that tells me “Not Enough Cases To Perform Cluster Analysis” – “Execution of this command stops”. The problem is that I do have plenty of cases; a total weighted sample of N=3,268 to be exact. Now, some of the variables do have missing values because of the skip pattern in the survey. But there are still plenty of cases, or so I would think.

Out of the 12 variables in one of the cluster analyses I have the following number of cases for each variable:

·         Q21_1 à Number of valid cases is 672

·         Q21_2 à Number of valid cases is 746

·         Q21_3 à Number of valid cases is 1833

·         Q21_4 à Number of valid cases is 2265

·         Q21_5 à Number of valid cases is 2414

·         Q21_6 à Number of valid cases is 2597

·         Q21_7 à Number of valid cases is 2704

·         Q21_8 à Number of valid cases is 1413

·         Q21_9 à Number of valid cases is 1423

·         Q21_10 à Number of valid cases is 802

·         Q21_11 à Number of valid cases is 2404

·         Q21_12 à Number of valid cases is 1610

My syntax is below:

QUICK CLUSTER q21_1 q21_2 q21_3 q21_4 q21_5 q21_6 q21_7 q21_8 q21_9 q21_10

    q21_11 q21_12

  /MISSING=PAIRWISE

/CRITERIA=CLUSTER(5) MXITER(100) CONVERGE(0)

/METHOD=KMEANS(NOUPDATE)

/PRINT INITIAL.

I don’t understand why it is doing this or if there is anything I can do to get it to run the cluster analysis. I would love to hear any thoughts.

Thank you in advance,

Colin

If you reply to this email, your message will be added to the discussion below:
http://spssx-discussion.1045642.n5.nabble.com/Not-Enough-Cases-To-Perform-Cluster-Analysis-tp5722164.html

To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

Art Kendall
Social Research Consultants