[SciPy-User] how does scipy.stats.t.isf work?
Warren Weckesser
warren.weckesser@enthought....
Wed Jun 16 11:42:06 CDT 2010
josef.pktd@gmail.com wrote:
> On Wed, Jun 16, 2010 at 12:13 PM, neurino <neurino@gmail.com> wrote:
>
>> Honestly the full question is
>> "how does scipy.stats.t.isf work compared to common spreadsheets function
>> T.INV?"
>> I have to translate an excel calculation using TINV
>> In all excel, ooo or abiword I get:
>>
>> TINV: inverse of the survival function of the Student t-distribution
>> Arguments:
>> p: probability
>> dof: number of degrees of freedom
>> The survival function is 1 minus the cumulative distribution function.
>> Note: If p < 0 or p > 1 or dof < 1 this function returns a #NUM! error.
>> Microsoft Excel: This function is Excel compatible.
>> Examples:
>> tinv(0,4;32) evaluates to 0,852998453651888.
>>
>> while with scipy I get:
>>
>>>>> from scipy.stats import t
>>>>> t.isf(.4, 32)
>>>>>
>> 0.25546356665122449
>> Any advice welcome, please consider I'm an informatic but not a
>> mathematician.
>> Thanks for your support
>>
>
> I guess Excel uses a two-sided tail probability
>
>
>>>> stats.t.isf(.4/2., 32)
>>>>
> 0.85299845247181938
>
>
Yes.
Check out the documentation for TINV here:
http://support.microsoft.com/kb/828340
Note that TINV(p, df) is the inverse for TDIST(x, df, 2). That '2' means
TDIST is two-sided. To quote from the above link:
"For any particular positive value of x, TDIST(x, df, 2) returns the
probability that a t-distributed random variable with df degrees of
freedom is greater than or equal to x or is less than or equal to –x."
So you will need to divide the probability by 2 to compare t.isf to TINV.
For example, this matches TINV(0.4; 32):
>>> t.isf(0.2, 32)
0.8529984524718196
Warren
> Josef
>
>
>
>> _______________________________________________
>> SciPy-User mailing list
>> SciPy-User@scipy.org
>> http://mail.scipy.org/mailman/listinfo/scipy-user
>>
>>
>>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
More information about the SciPy-User
mailing list