Python Programming for Biology: Bioinformatics and Beyond

Two-sample T-test

The two-sample T-test tests whether two independent samples have the same underlying,

true mean. In R there is a method, ‘t.test’, to carry this out. There are a few extra subtleties

in comparison with the RPy code for the binomial test.

The T-test involves comparing two collections of samples. These need to be converted

to  RPy  data  types  using  the  converter  R.FloatVector().  The  R  method,  ‘t.test’,  has  an

optional argument, var.equal,  which  specifies  whether  the  two  samples  should  be  treated

as  having  the  same  variance  or  not.  In  Python,  var.equal  cannot  be  used  as  a  key  in  a

function argument because it is not valid syntax. In RPy we can often use these arguments

by changing the dot (.) to an underscore (_). But replacing var.equal with var_equal does

not  work  here  and  instead  we  use  a  bit  of  Python  trickery.  In  Python  you  can  specify

‘key=value’ arguments by using a dictionary. But the trickery is that the dictionary itself is

not  passed  in  (the  function  is  not  expecting  a  dictionary  as  an  argument)  but  instead  we

use  a  magic  **  in  front  of  the  dictionary,  which  means  that  the  dictionary  contents  are

unwrapped  into  key=value  pairs.  This  is  the  safest  way  to  pass  arguments  with  dots  in

them in RPy.

We  pass  back  both  the  T  statistic,  which  happens  to  be  item  0,  and  the  probability,

which happens to be item 2, of the result.

def tTest(x, y, sameVariance=False):

func = R.r['t.test']

argDict = {'var.equal': sameVariance}

---

result = func(x=R.FloatVector(x), y=R.FloatVector(y), **argDict)

return result[0][0], result[2][0]

We can now test the function.

from numpy import array

samples1 = array([1.752, 1.818, 1.597, 1.697, 1.644, 1.593])

samples2 = array([1.878, 1.648, 1.819, 1.794, 1.745, 1.827])

print('Same variance result', tTest(samples1, samples2, sameVariance=True))

# Result is: -2.072, 0.0650

print('Not same variance result', tTest(samples1, samples2,

sameVariance=False))

# Result is: # -2.072 0.0654