| T O P I C R E V I E W |
| OndrejMM |
Posted - 07/02/2008 : 12:00:01 AM
Hi Guys,
I have a specific problem with “statistic on column”,… I think the Origin first assumes that data have normal distribution and then do the statistic on a column,… but if you check my data (see histogram below) it is apparently not normal distribution,…
so my question is whether is it possible to determine the type of distribution of the experimental data first and then do statistisc on the column with respect to determined data distribution,…???
Many thanks for your help.
Ondrej
|
| 4 L A T E S T R E P L I E S (Newest First) |
| OndrejMM |
Posted - 07/04/2008 : 12:23:08 AM
Hi Easwar,
Thanks a lot for the extensive explanation. I was puzzling what those all statistical parameters mean,… and I wasn't sure how the Origin does that statistical calculatios.
Many Thanks
Ondrej
PS: I still cannot insert an image.
|
| easwar |
Posted - 07/02/2008 : 9:33:57 PM
Hi Ondrej,
The ability to upload images to the forum was broken and has been fixed. So you can try again to upload your image.
As Echo said, none of the quantities computed and reported in descriptive stats in Origin assumes any underlying distribution. For example, the mean is simply the arithmetic mean (sum of all elements divided by the number of elements).
What matters is how the results are interpreted and what informatio do the results provide, as to the nature of the underlying distribution. For example, skewness would be zero for a perfectly symmetrical distribution. A symmetrical distribution does not necessarily have to be a normal distribution by the way. So if sknewness is different from zero, one can interpret that the underlying distribution is not symmetrical. Kurtosis, which measures the peakedness of the distribution, would be zero for a perfectly normal distribution. So a value larger or smaller than zero means the distribution is more, or less peaked than a normal distribution. So the meaning of Kurtosis is tied to normal distribution, unlike skewness.
So some of the results have specific interpretation based on a normal distribution, and others do not. But again none of the computaions assume anything about the underlying distribution.
Easwar
OriginLab
|
| OndrejMM |
Posted - 07/02/2008 : 6:31:24 PM
Hi Echo,
Thank you for your response,… I wanted to add my histogram but for some reason I cannot insert a image,…
anyway I really want to do only the statistic on column, but in respect to the data distribution,… I assume that e.g. “the mean” will be different if I apply normal distribution or if I apply let say lognormal distribution,… what do you think? I’m not really sure now,…
so, I would say that I need to first determine type of data distribution and then do the statistic in respect to determined distribution,…
anyway I will keep thinking,…
many thanks
Ondrej
|
| Echo_Chu |
Posted - 07/02/2008 : 5:08:55 PM
Hi, Ondrej
Statistics on column is not based on any distribution assumption. But some result such as mean, skewness and kutosis make implicit assumption about the underlying distribution.
I am not sure what you want to do with you data next, if hypothesis test or ANOVA, such analysis require data are from normal distribution. I will suggest you use Non-parameter analysis insteadly
Echo
OriginLab Corp |
|
|
