Crystal Reports Standard Deviation

[cdj01]

I use Crystal 8 connected to an Oracle database to analyze laboratory results. I have come to the conclusion that the standard deviation function is flawed. It seems that the larger our sample population, the farther off the standard deviation becomes, which does not fit our data. To confirm this, I exported my crystal data to Excel, and used Excel's standard deviation function. Crystal was reporting a stdev of 1.21, whereas Excel reported 0.22. This was off of the same set of data.
Is there a limit to the number of samples crystal can accurately perform calculations on? I am running my crystal query against on year's worth of data, about 266,000 results. Both crystal

 

[LidoDeJesolo]

It seems we only have half of your post...

Lido
Development & Reporting
UK

 

[cdj01]

No actually that was it, I missed a couple of words when I trimmed something. Basically it's all there, using the same set of data with Excel, I get a drastically different standard deviation than when I compute it in Crystal 8.

 

[Reebo99]

Crystal has 2 Standard Deviation summary functions :

StdDev and
PopulationStdDev

Do you get the same result using both functions?



Reebo
UK

Please Note - Due to current economic forcast and budget constraints, the light at the end of the tunnel has been switched off. Thank you for your cooperation.

 

[cdj01]

I got excited, not knowing about the population stdev (I know nothing about statistics, I was just asked for the standard deviation) but unfortunately the results were the same or very close to it when I ran a populationstdev.

 

[Reebo99]

That's buggered up my idea then. Sorry I can't be of more help.

Reebo
UK

Please Note - Due to current economic forcast and budget constraints, the light at the end of the tunnel has been switched off. Thank you for your cooperation.

 

[mbarron]

Shot in the dark time.

Are you suppressing any data in Crystal? Crystal evaluates the data even if you can't see it. When you export and evaluate in Excel, you would be doing the StdDev on only what you can see.

Mike

 

[cdj01]

I don't have anything set to be supressed. It's a very simple report. It is basically a straight table pull (of the table RESULT_VIEW) with the following in the selection expert:
{RESULT_VIEW.MNEMONIC}='WAL' AND
{RESULT_VIEW.TESTSEQ}='1' AND
{RESULT_VIEW.TESTCODE} IN ['06004','05004','02004','11001'] AND
{RESULT_VIEW.LABNO} IN '20030101001' TO '20040109999''

A quick description of what I am doing: This is a laboratory results database. 'mnemonic' is the test result, here I want normal results hence selecting 'WAL' which is our normal result. The test may be performed multiple times, and I only want the initial result, which is TESTSEQ 1. There are 4 different tests I want here, hence the 4 TESTCODEs. And the LABNO is the range of specimens I want, 1 years worth.
I have the results grouped by TESTCODE, and I have a standard deviation for each group with the code StdDev ({RESULT_VIEW.VALUE},{RESULT_VIEW.TESTCODE})
VALUE being the measurement of the analyte that we base our results on.