Descriptives discrepancy in Sums and Means
Locked
 
|
Can anyone explain this apparent discrepancy in the sums and means from this Descriptives output?
For example, note the results for v14, v15, and v16. I'm using SPSS 21. Descriptive Statistics N Sum Mean v12 8 3 .38 v13 8 2 .25 v14 8 1 .13 v15 8 0 .00 v16 8 0 .13 v17 8 2 .25 v18 8 3 .38 v23 8 3 .38 v24 8 2 .25 v25 8 1 .13 v26 8 0 .00 v27 8 0 .13 v28 8 2 .25 v34 8 3 .38 v35 8 2 .25 v36 8 1 .13 v37 8 0 .00 v38 8 0 .13 v45 8 3 .38 v46 8 2 .25 v47 8 1 .13 v48 8 0 .00 v56 8 3 .38 v57 8 2 .25 v58 8 1 .13 v67 8 3 .38 v68 8 2 .25 v78 8 3 .38 Valid N (listwise) 8 |
Re: Descriptives discrepancy in Sums and Means
Administrator
|
Odd! Maybe attach the raw data to this thread.
http://spssx-discussion.1045642.n5.nabble.com/template/NamlServlet.jtp?macro=reply&node=5724728
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
Re: Descriptives discrepancy in Sums and Means
Administrator
|
Or if it is easier for you (and given that n=8), list the data for variables V16, V27 and V38 (which appear to be the problematic ones), and then paste the listed data into your next post to the list. Here's the LIST command:
LIST V16 V27 V38. HTH.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
Re: Descriptives discrepancy in Sums and Means
|
Administrator
|
I predict that we will not replicate it with any subset of pasted data.
There is a peculiarity to the output which suggests something really weird going on. Note the weird ones have a sum of 1 -> mean .13, followed by a sum of 0 with a correct mean of 0 then another 0 with an incorrect mean of .13 Really weird. Can't be explained by weighting or rounding. --
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
Re: Descriptives discrepancy in Sums and Means
|
In reply to this post by olsenja
are any of you values
negative for V16 V27 V38.
if you post your data, please use this syntax. formats V16 V27 V38 (f20.16). LIST V16 V27 V38. Art Kendall Social Research ConsultantsOn 3/4/2014 4:35 PM, olsenja [via SPSSX Discussion] wrote: Can anyone explain this apparent discrepancy in the sums and means from this Descriptives output?
Art Kendall
Social Research Consultants |
Re: Descriptives discrepancy in Sums and Means
|
|
In reply to this post by olsenja
Two of these are easy to explain:
1. V14 has a mean of .13, which is the rounded version of the actual mean (1/8) of .125. So, this is likely not a problem. (Just like V25, V58, etc.) 2. V15 has a sum and mean of 0. I do not see a problem with this one. Not sure about V16. What are the actual data values. But, since there are only eight of them, you could calculate by hand and then see what is happening. I also would suggest that you include other output (e.g., min and max values and SD) which are useful in trying to figure out apparent issues and difficulties from the data. Harley Dr. Harley Baker Professor of Psychology Internal Evaluator, Project ACCESO Madera Hall 2413 California State University Channel Islands One University Drive Camarillo, CA 93012 805.437.8997 (p) 805.437.8951 (f) [hidden email] ________________________________________ From: SPSSX(r) Discussion [[hidden email]] on behalf of olsenja [[hidden email]] Sent: Tuesday, March 04, 2014 1:35 PM To: [hidden email] Subject: Descriptives discrepancy in Sums and Means Can anyone explain this apparent discrepancy in the sums and means from this Descriptives output? For example, note the results for v14, v15, and v16. I'm using SPSS 21. Descriptive Statistics N Sum Mean v12 8 3 .38 v13 8 2 .25 v14 8 1 .13 v15 8 0 .00 v16 8 0 .13 v17 8 2 .25 v18 8 3 .38 v23 8 3 .38 v24 8 2 .25 v25 8 1 .13 v26 8 0 .00 v27 8 0 .13 v28 8 2 .25 v34 8 3 .38 v35 8 2 .25 v36 8 1 .13 v37 8 0 .00 v38 8 0 .13 v45 8 3 .38 v46 8 2 .25 v47 8 1 .13 v48 8 0 .00 v56 8 3 .38 v57 8 2 .25 v58 8 1 .13 v67 8 3 .38 v68 8 2 .25 v78 8 3 .38 Valid N (listwise) 8 -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Descriptives-discrepancy-in-Sums-and-Means-tp5724728.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 |
Re: Descriptives discrepancy in Sums and Means
|
In reply to this post by olsenja
At 04:35 PM 3/4/2014, olsenja wrote:
>Can anyone explain this apparent discrepancy in the sums and means >from this Descriptives output? For example, note the results for >v14, v15, and v16. > >Descriptive Statistics > N Sum Mean >v12 8 3 .38 >v13 8 2 .25 >v14 8 1 .13 >v15 8 0 .00 >v16 8 0 .13 ?? >v17 8 2 .25 >v18 8 3 .38 >v23 8 3 .38 >v24 8 2 .25 >v25 8 1 .13 >v26 8 0 .00 >v27 8 0 .13 ?? >v28 8 2 .25 >v34 8 3 .38 >v35 8 2 .25 >v36 8 1 .13 >v37 8 0 .00 >v38 8 0 .13 ?? >v45 8 3 .38 >v46 8 2 .25 >v47 8 1 .13 >v48 8 0 .00 >v56 8 3 .38 >v57 8 2 .25 >v58 8 1 .13 >v67 8 3 .38 >v68 8 2 .25 >v78 8 3 .38 >Valid N (listwise) 8 As Harley Baker has just noted, all the means are the correct values of Sum/N, rounded to two decimal places, EXCEPT v16, v27,, and v38. For those three, Sum is 0, but Mean is the value that would be correct for Sum=1. There are several other variables (v15, v26, v37, and v48) for which Sum=0 and Mean has the correct value, .00. In all three cases where the mean appears to be wrong, Sum=0 for both the problem variable AND THE PRECEDING VARIABLE; and there are no instances of two successive variables with Sum=0 and the mean reported correctly for the second variable. These observations should contain some clues. How could the observed results happen? Easy: they couldn't. However, "What avails the classic bent And what the cultured word, Against the undoctored incident That actually occurred?" -Rudyard Kipling ===================== 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 |
|
|
In reply to this post by Baker, Harley
Here is an abbreviated descriptives output along with an extended precision listing of the original data. These are just ordinary dummy variables. There are no negative values. Min and Max are apparent from the data and the standard deviations are given, but don't inform the problem.
--Joe formats v14 v15 v16 (f20.16). Descriptive Statistics N Sum Mean Std. Deviation v14 8 1.0000000000001000 .125000000000100 .353553390593374 v15 8 .0000000000000000 .000000000000000 .000000000000000 v16 8 .0000000000000000 .125000000000100 .353553390593374 Valid N (listwise) 8 list v14 V15 V16. v14 v15 v16 1.0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 1.0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 .0000000000000000 Number of cases read: 8 Number of cases listed: 8
|
Descriptives discrepancy in Sums and Means
|
|
In reply to this post by Richard Ristow
In Statistics V21.0.0.0 an improvement
was made to the computation of the sum statistics to yield higher accuracy
with numbers that vary a lot in magnitude or very large sums (See http://en.wikipedia.org/wiki/Floating_point
for a detailed article on the wonders of floating point arithmetic). Unfortunately
in implementing the Kahan algorithm, which computes the sum in a roundabout
way that produces higher precision, a problem was introduced in a few situations.
This was repaired in a hot fix IIRC and then in FixPack 1 for V21.0.0.0.
It only affected the sum statistic in AGGREGATE, DESCRIPTIVES, and
FREQUENCIES and only for these special cases. No other procedures were
affected. SUMMARIZE, for example, continued to compute the sum correctly.
I believe that this issue was discussed on this list at the time,
but I have not gone back to search the archives.
Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] phone: 720-342-5621 |
Re: Descriptives discrepancy in Sums and Means
|
Administrator
|
There's some info here:
http://www-01.ibm.com/support/docview.wss?uid=swg27038029 Under "Statistical Analysis": A problem where AGGREGATE and DESCRIPTIVES could produce an incorrect sum with certain data (mostly 0s) was fixed.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |