From: Ajasja Ljubetič (ajasja.ljubetic_at_gmail.com)
Date: Thu Jun 26 2014 - 08:05:48 CDT
There are also a lot of normality tests
<http://en.wikipedia.org/wiki/Normality_test> Such as Anderson-Darling
<http://en.wikipedia.org/wiki/Anderson%E2%80%93Darling_test>, Shapiro-Wilk
<http://en.wikipedia.org/wiki/Shapiro%E2%80%93Wilk_test>, etc.. Be advised
that these only work well for small sample sizes (less than ~500-1000
samples). If very large samples are used the tests always reject
<http://stats.stackexchange.com/questions/2492/is-normality-testing-essentially-useless>
the distribution. The solution I've used in the past is just to randomly
draw 500 samples from my original huge sample set.
There are also Q-Q plots
<http://en.wikipedia.org/wiki/Quantile-quantile_plot> for graphically
determining deviation from normality (or in general the deviation of two
probability distributions).
Best regards,
Ajasja
On 25 June 2014 19:12, Giacomo Fiorin <giacomo.fiorin_at_gmail.com> wrote:
> Hello Mustafa, there are a multitude of recipes and opinions on this.
> Generally, you would simply calculate the 3rd and 4th statistical moments
> of the distribution, comparing them to those of the normal one (Gaussian).
>
> Giacomo
>
>
> On Wed, Jun 25, 2014 at 6:50 AM, Mustafa Tekpinar <tekpinar_at_buffalo.edu>
> wrote:
>
>> Hi,
>> I am performing SMD simulations and I want to calculate Potential of Mean
>> Force (PMF) using those simulations. As far as I know, work distribution
>> (or force distribution) has to be gaussian to calculate PMF (with
>> Jarzynski's method and second order cumulant expansion). I was wondering
>> how can I check if my work distribution is gaussian.
>>
>> Thanks in advance,
>> Mustafa Tekpinar
>>
>
>
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:33 CST