Re: understanding total for standard deviation

The "total" mean and standard deviation are calculated by considering
the set of all of the measured test times. Unless your tests are similar
in nature, neither the total mean or total standard deviation is likely
to be of practical use.

[If you're of a statistical bent, you may also like to know that all of
the standard deviation values use the population standard deviation, not
the sample standard deviation. See
http://en.wikipedia.org/wiki/Bessel%27s_correction. The sample sd is
more appropriate, and I may fix it one day, but as a practical way of
presenting variability in the tests, either figure suffices].

- Phil

On 03/04/12 15:53, Kees Hink wrote:
> I don't understand the calculation of the "total" value for Standard
> Deviation (SD) as displayed in the console and logfile.
>
> I have 7 tests resulting in the log below [1]. I understand that
> instead of a total, i shows the mean value for MT and MRL. But how
> does the Grinder arrive at 2315.26 for total SD? I read the FAQ,
> notably [2], but found no answer there.
>
> [1] CSV logfile output:
> ID	# T	# E	MT (ms)	SD (MS)	MRL (bytes)	# RE	Name
> Test 101	10	0	19.1	16.35	43972	0	Frontpage, Anon
> Test 102	10	0	38.2	62.58	227489	0	Many comments, Anon
> Test 103	10	0	9.5	0.5	25572	0	No comments, Anon
> Test 110	10	0	99.1	3.08	0	0	POST login
> Test 111	10	0	1038.2	746.64	45799	0	Frontpage, Auth
> Test 112	10	0	6744.8	14.15	264224	0	Many comments, Auth
> Test 113	10	0	294.4	2.5	32048	0	No comments, Auth
> Totals	70	0	1177.61	2315.26	91300.57	0	
>
> [2] http://grinder.sourceforge.net/faq.html#totals-and-composite-tests
>

------------------------------------------------------------------------------
Better than sec? Nothing is better than sec when it comes to
monitoring Big Data applications. Try Boundary one-second 
resolution app monitoring today. Free.
http://p.sf.net/sfu/Boundary-dev2dev