Community Server

Can statistics101 do this?

That is, given a sample size (or degrees of freedom) and a proportion of the distribution of interest (say .05 of the right tail) can statistics101 give me a t-value or an f-value?

For example, in R I can tell the program that I want to use the t-distribution and that I 98 degrees of freedom and I want a one-tailed test at the .05 level and R will return the associated t-value.

Thanks,

Sherman

Sherman,

You can use the subroutine TDIST to generate a t distribution, then use PERCENTILE to find the t value, remembering that for the right tail, you'd need to use 100 minus the desired proportion. In Resampling Stats code:

INCLUDE "lib\tDistribution.txt"
'. . .
TDIST 10000 98 tDistribution                 'Draw numbers from the t distribution.
PERCENTILE tDistribution 95 t                'Find t for p = 0.05.
HISTOGRAM percent binsize 0.25 tDistribution 'Histogram added for visualization
PRINT t

But a t distribution with 98 degrees of freedom is effectively a normal distribution.

I haven't yet added an f distribution to the library, but here's a subroutine to generate one if you need it:

INCLUDE "lib\chiSquareDistribution.txt"
'. . .
NEWCMD FDIST size df1 df2 result
   CHISQUAREDIST size df1 X
   CHISQUAREDIST size df2 Y
   LET result = (X/df1) / (Y / df2)
END

where "size" is the number of values to generate from the f distribution with degrees of freedom "df1" and "df2".

regards,

John

Hi John,

This is exactly what I was looking for...I didn't realize there was already a library that included the t-distribution, etc. Thanks for your help once again!

Sherman