SPSS Warnings

[StatsUser]

CONTENTS DELETED

The author has deleted this message.

[David Marso]

I believe you need to post the text of the warning!

StatsUser wrote

Hello,

I am analyzing data using a split file, followed by a paired t test. I get the warning below.

SPSS gives the warning but calculates the statistic anyway. I cross checked with turning off the split file (descriptives, explore) – the means, SDs, and dfs were the same as the paired t tests output. I also googled the warning and came across some output where the warnings were present but the paired t test was not calculated.

Any clarity on this would be greatly appreciated.

Can I still report these statistic despite the warning given that spss calculated it anyway and the means/SDs/DFs are the same when I turn off split file (descriptives, explore)?

Please help.

Thanks,

-ps. I need get a significance value hence the paired t test vs just doing descriptives, explore

[StatsUser]

CONTENTS DELETED

The author has deleted this message.

[Art Kendall]

try a crosstab of the variables that are used to split the files.
Are any of the values missing?

also should you really by doing a 2 or 3 way anova with between factors whatever you are splitting the file on and a with factor for the repeats?
On 7/26/2012 1:50 PM, StatsUser wrote:

Yes, thank you.

Here is the warning:


Warnings
No statistics are computed for a split file in the Paired Samples
Correlations table. The split file is: Violation Manipulation =..
No statistics are computed for a split file in the Paired Samples Test
table. The split file is: Violation Manipulation =..





--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/SPSS-Warnings-tp5714472p5714475.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

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

[StatsUser]

CONTENTS DELETED

The author has deleted this message.

[Rich Ulrich]

I think you should take a closer look at that Frequency.

As I read your error message, your Split File is finding
*three* groups,  "1", "2", and "." (system missing).
I am guessing that your file has a case (probably just 1)
that has every value as missing, because you read in or
imported a blank line.

The error message did not affect the statistics on the
two "splits" that did have valid data.

--
Rich Ulrich

> Date: Thu, 26 Jul 2012 12:07:10 -0700

> From: [hidden email]
> Subject: Re: SPSS Warnings
> To: [hidden email]
>
> Hi,
>
> Yes I did a frequency and nothing is missing (it is one categorical
> variable, 1=male, 2=female, that I have split filed)
>
> I’m not sure what you mean by a 2 or 3 way anova. It is one categorical
> variable with gender being split filed, and DV is perceptions of driving
> behavior pre and post manipulation. I would like to see if there are sig
> diffs pre and post DV for women and men separately (not together)
>
> 

...

[David Marso]

You are obviously missing the point of split file!
Tell us about "Violation Manipulation "  Your warning suggests that it has one or more system missing values and together those missing cases have 0 variance or some other anomaly prohibiting calculation of the stats.
Perhaps post your syntax from the point of
SORT CASES BY blah blah blah.
SPLIT FILE BY blah blah blah.
T-TEST SOMETHING OTHER THAN blah blah blah.
---

StatsUser wrote

Hi,

Yes I did a frequency and nothing is missing (it is one categorical variable, 1=male, 2=female, that I have split filed)

I’m not sure what you mean by a 2 or 3 way anova. It is one categorical variable with gender being split filed, and DV is perceptions of driving behavior pre and post manipulation. I would like to see if there are sig diffs pre and post DV for women and men separately (not together)

[StatsUser]

CONTENTS DELETED

The author has deleted this message.

[Bruce Weaver]

But taking up Art's point, why not use a two-factor ANOVA model, with Gender as a between-subjects factor, and time (pre & post) as a within-subjects (or repeated measures) factor?  That model will let you test the main effects of Gender and Time as well as their interaction.  The null hypothesis for the interaction term is that the pre-post difference (which seems to be your main interest) is the same in both groups.  If you get a statistically significant F-test for the interaction, you can reject that null hypothesis.  

The most common way to run that model is via Analyze - GLM - Repeated Measures.  You have to enter some info for the repeated measures variable in the first dialog that comes up.  Name the within-Ss factor Time, and indicate that there are 2 levels, then click ADD followed by Define.  

In the next dialog that appears, move your pre and post variables into the appropriate spots in the top box, and add gender to the Between-S factors box.  (It probably has to be a numeric variable, so if you have it as a string, you'll have to recode it first.)  

A plot is useful in helping you understand the nature of the interaction (if there is one).  You want Time as the horizontal axis with separate lines for Gender.

You probably want the "estimated marginal means" too.  Click on Options to bring up the dialog that allows you to select those.  The most common approach is to select all of them except the Overall, which is usually of less interest than the others.  

Exit the main dialog via the PASTE button, and you should see some syntax that looks something like this:

GLM pre post BY gender
  /WSFACTOR=Time 2 Polynomial
  /METHOD=SSTYPE(3)
  /PLOT=PROFILE(Time*gender)
  /EMMEANS=TABLES(gender)
  /EMMEANS=TABLES(Time)
  /EMMEANS=TABLES(gender*Time)
  /CRITERIA=ALPHA(.05)
  /WSDESIGN=Time
  /DESIGN=gender.

Highlight it and run it (via one of the Run options on the toolbar, or by CTL-R).

In the output, the main effect of Time and the Time*Gender interaction will appear in the Tests of Within-Ss Effects box, and the main effect of Gender will appear in the Tests of Between-Ss Effects box.  (I might not have those labels exactly right, but they're close.)  

Finally, you should be able to find lots of online tutorials on how to run this model.  Try searching on <SPSS mixed design ANOVA> or <SPSS split-plot ANOVA> or <SPSS between-within ANOVA>, for example.

HTH.

StatsUser wrote

Thank you Rich (and all who took time to respond),


That was the problem!  I no longer get the warning message, many thanks for your help.

[David Marso]

I concur with Bruce.  This is a much more appropriate and informative analysis.

Bruce Weaver wrote

But taking up Art's point, why not use a two-factor ANOVA model, with Gender as a between-subjects factor, and time (pre & post) as a within-subjects (or repeated measures) factor?  That model will let you test the main effects of Gender and Time as well as their interaction.  The null hypothesis for the interaction term is that the pre-post difference (which seems to be your main interest) is the same in both groups.  If you get a statistically significant F-test for the interaction, you can reject that null hypothesis.  

The most common way to run that model is via Analyze - GLM - Repeated Measures.  You have to enter some info for the repeated measures variable in the first dialog that comes up.  Name the within-Ss factor Time, and indicate that there are 2 levels, then click ADD followed by Define.  

In the next dialog that appears, move your pre and post variables into the appropriate spots in the top box, and add gender to the Between-S factors box.  (It probably has to be a numeric variable, so if you have it as a string, you'll have to recode it first.)  

A plot is useful in helping you understand the nature of the interaction (if there is one).  You want Time as the horizontal axis with separate lines for Gender.

You probably want the "estimated marginal means" too.  Click on Options to bring up the dialog that allows you to select those.  The most common approach is to select all of them except the Overall, which is usually of less interest than the others.  

Exit the main dialog via the PASTE button, and you should see some syntax that looks something like this:

GLM pre post BY gender
  /WSFACTOR=Time 2 Polynomial
  /METHOD=SSTYPE(3)
  /PLOT=PROFILE(Time*gender)
  /EMMEANS=TABLES(gender)
  /EMMEANS=TABLES(Time)
  /EMMEANS=TABLES(gender*Time)
  /CRITERIA=ALPHA(.05)
  /WSDESIGN=Time
  /DESIGN=gender.

Highlight it and run it (via one of the Run options on the toolbar, or by CTL-R).

In the output, the main effect of Time and the Time*Gender interaction will appear in the Tests of Within-Ss Effects box, and the main effect of Gender will appear in the Tests of Between-Ss Effects box.  (I might not have those labels exactly right, but they're close.)  

Finally, you should be able to find lots of online tutorials on how to run this model.  Try searching on <SPSS mixed design ANOVA> or <SPSS split-plot ANOVA> or <SPSS between-within ANOVA>, for example.

HTH.

StatsUser wrote

Thank you Rich (and all who took time to respond),


That was the problem!  I no longer get the warning message, many thanks for your help.

[Art Kendall]

But would it not be worthwhile to see if there is a difference in the
changes from pre to post for males and females?
Also, if you did a two way anova instead of 2 one-way anova you would
have more power to detect differences.
(A t-test is an anova with 1 degree of freedom. )

Check and see if you had done a previous split file by manipulation and
violation.
If so,
split file off.
sort cases by gender.
split file by gender.

look at the error message from this example syntax.

open a new instance of SPSS.
paste the syntax below into a new syntax window.
run it.

data list list /id(f2) gender(f1) pre(f2) post(f2).
begin data
1 1 3 5
2 1 2 5
3 1 4 5
4 2 1 4
5 2 2 3
6 2 4 5
7 . 3 5
end data.
sort case by gender.
split file by gender.
t-test pairs = pre with post.


When I see your previous error message I wonder if you should have more
the one between subjects factor.


Art Kendall
Social Research Consultants

On 7/26/2012 3:04 PM, Gill, Ritu wrote:

>
> Hi,
>
> Yes I did a frequency and nothing is missing (it is one categorical
> variable, 1=male, 2=female, that I have split filed)
>
> I’m not sure what you mean by a 2 or 3 way anova. It is one
> categorical variable with gender being split filed, and DV is
> perceptions of driving behavior pre and post manipulation. I would
> like to see if there are sig diffs pre and post DV for women and men
> separately (not together)
>
> Ritu
>
> *From:*Art Kendall [mailto:[hidden email]]
> *Sent:* July-26-12 2:58 PM
> *To:* Gill, Ritu
> *Cc:* [hidden email]
> *Subject:* Re: [SPSSX-L] SPSS Warnings
>
> try a crosstab of the variables that are used to split the files.
> Are any of the values missing?
>
> also should you really by doing a 2 or 3 way anova with between
> factors whatever you are splitting the file on and a with factor for
> the repeats?
>
> Art Kendall
> Social Research Consultants
>
> On 7/26/2012 1:50 PM, StatsUser wrote:
>
>     Yes, thank you.
>
>
>
>     Here is the warning:
>
>
>
>
>
>     Warnings
>
>     No statistics are computed for a split file in the Paired Samples
>
>     Correlations table. The split file is: Violation Manipulation =..
>
>     No statistics are computed for a split file in the Paired Samples Test
>
>     table. The split file is: Violation Manipulation =..
>
>
>
>
>
>
>
>
>
>
>
>     --
>
>     View this message in context:http://spssx-discussion.1045642.n5.nabble.com/SPSS-Warnings-tp5714472p5714475.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]  <mailto:[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

[StatsUser]

CONTENTS DELETED

The author has deleted this message.

[Art Kendall]

What did it turn out to be?
On 7/26/2012 8:34 PM, StatsUser wrote:

Thanks everyone! Problem solved!



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/SPSS-Warnings-tp5714472p5714493.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

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

[StatsUser]

CONTENTS DELETED

The author has deleted this message.