Re: Error with descriptive statistics?

Posted by Rich Ulrich on Apr 13, 2013; 2:31am
URL: http://spssx-discussion.165.s1.nabble.com/Error-with-descriptive-statistics-tp5719435p5719444.html

Your data orientation does not seem to be primarily to the "data base"
organization of SPSS.  I am thinking especially of your (3) -- Excel will allow
you to treat a cell as text so you can enter two numbers.  But an SPSS
"variable" is always  "a single response,"  so this comment is wholly out
of place.

But I also notice that you say that you "only paired the relevant variables."
That is also an inappropriate (for SPSS) description.  For SPSS, you would
say that you used KEEP  to save only the relevant 8 variables.   And you should
also keep an identifier for which Year, so that rows are identified in the new set.
If Year is not in the file, you can use /IN= IN2012, and so on.

With the IN variables, you can use a Filter on the combined file and recover
(presumably) the original counts that exist in the separate files.

What you report in (4) shows that you are really messing up in some way.
One thing that I imagine going wrong is that you failed to ADD CASES and
instead you merge-matched on ID.  That would give you a file with about N instead
of 2N  cases -- assuming a match for most records.  In that new file, when the
cases matched on ID, the data would be used for the 1st file named for all the
variables that exist in the first file.  On ADD CASES, SPSS does *not*  "pair up"
any values; your new count should be exactly the total of the two separate counts.

As David says, we can tell more about what went wrong if you show us what
you actually did.  - Cut-and-paste for us the syntax that you actually ran.
If you use the GUI interface ... you do need to PASTE the syntax that you are
running, for documentation for yourself, and to have it available to copy to us.

--
Rich Ulrich

---

Date: Fri, 12 Apr 2013 23:33:04 +0000
From: [hidden email]
Subject: Re: Error with descriptive statistics?
To: [hidden email]

Thanks all for questions. Here’s some more info:

 

1.       The error occurs only occasionally. I noticed it because I have a data set with a large number of variables (survey data). I only need to analyze a few of the variables and need to compare them to data from a previous year’s survey. To compare the surveys, I merged the using Date>Merge>Add Cases. Since the surveys asked different questions, and most are irrelevant to my task at hand, I only paired the relevant variables (8 out of 100+). Then I ran descriptives on the merged dataset and the original datasets, and the results were different (I expected that the two datasets would add to the results of the merged one). That led me to just delete the unnecessary variables from one of the original datasets and run descriptives. When I did that, the results were different again. That lead me to copy and paste the data from the original dataset and the dataset with fewer variables (but the same cases) into excel and run a simple countif function on them. That’s how I figured out that the data was identical, but the descriptives being produced by SPSS were different.

 

2.       The variable I’m looking at is numeric with 0 decimal places. The variable is set to a nominal scale, because the numbers represent transportation mode (1=drive alone, 2=carpool, etc.) Zero (0) is not a value for the variable – that is, none of my cases have 0 as the answer to the question. None of the data have decimals.

3.       The question only allows for a single response. (So each case just has one number in the data field for this variable.)

4.       There are user-defined missing values in my file. And that number changes in the descriptive statistics between the datasets. The total number of cases is constant between the two datasets.

5.       Filters aren’t turned on.

6.       Almost all the frequencies differ between the two datasets  – unfortunately the error isn’t limited to a single option within the variable (such as Drive Alone trips).

 

Any solutions spring to mind?

 

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Zdaniuk, Bozena
Sent: Friday, April 12, 2013 3:50 PM
To: [hidden email]
Subject: Re: Error with descriptive statistics?

 

the Freq table will report a value twice if the variable is set to no decimal points but there is a value=0 and a value=0.02, for example.
bozena

---

From: SPSSX(r) Discussion [[hidden email]] on behalf of R B [[hidden email]]
Sent: Friday, April 12, 2013 3:33 PM
To: [hidden email]
Subject: Re: Error with descriptive statistics?

I encountered something similar a long time ago with an older version. I obtained frequencies associated with the value zero twice in the frequency distribution Table. I can't recall how it was resolved. But I haven't experienced any problems with descriptives for several years now. My guess is that it is user-error. Why don't you open a new SPSS data file, enter some data, and see if the error continues to occur.

 

Ryan

 

On Fri, Apr 12, 2013 at 6:18 PM, Michelle D. Zeidman <[hidden email]> wrote:

Hi there,
 
I am running frequencies in descriptive statistics in SPSS 21 and the results are wrong. That is, it reports inaccurate counts of each response type to a survey question in the output table. Has anyone else experienced this? Your guidance on what I may be doing wrong would be greatly appreciated!
 
Thanks,
Michelle

 

Michelle Zeidman

Transit Program Operations Specialist  | UW Transportation Services

Office: 206-616-6087 | Cell: 206-518-1490 | [hidden email]