7 Replies Latest reply: Jan 28, 2013 7:05 AM by Hiroto
Level 1 (0 points)

hi ! i discovered statistical functions in numbers, and did a t-test

but I'd be very interested in calculating also the statistical significance, as a "p-value" (or alpha risk)

is there a way in Numbers, or adding a plug-in?

thanks a lot

Numbers iWork, OS X Mountain Lion (10.8.2)
• Level 7 (29,945 points)

A,

TDIST is used in Excel to calculate p-values, and the same function is included in Numbers. Use the Function Browser to review the syntax for the function:

The TDIST function returns the probability from the student’s t-distribution.

• TDIST(non-neg-x-value, degrees-freedom, tails)
• non-neg-x-value:  The value at which you want to evaluate the function. non-neg-x-value is a number value that must be greater than or equal to 0.
• degrees-freedom:  Degrees of freedom. degrees-freedom is a number value and must be greater than or equal to 1.
• tails:  The number of tails to return.
• one tail (1)Return the value for a one-tailed distribution.
• two tails (2)Return the value for a two-tailed distribution.

Jerry

• Level 5 (6,030 points)

Hello

As far as I know, TTEST() function returns the p-value of the specified t-test.

Regards,

H

• Level 1 (0 points)

jerrold,

thank you very much, it seems working  with TDIST, which in french is LOI.STUDENT

thanks a lot!

hiroto,

isn't that the t value with t-test only?

• Level 1 (0 points)

actually, the ttest function might return the pvalue. in that case, how can I have my t value???

thanks

• Level 5 (6,030 points)

Hello

The maps between t value and p value are provided by TINV() and TDIST().

```t value = TINV(p value, degrees of freedom)
p value = TDIST(t value, degrees of freedom, number of tails)
```

The t distribution is determined by the degrees of freedom (d.o.f.) of the t statistic in use and the d.o.f. is determined by the type of t-test and the sample size.

E.g.

For t-test of paired samples X (sample size = N, i.e., pair count = N, data count = 2 * N), d.o.f. = N - 1.

For t-test of two independent samples X1, X2 assumed to have equal population variance (sample size = N1, N2 resp.), d.o.f. = N1 + N2 - 2.

If you want t value, you may a) use t = TINV(TTEST(), d.o.f.) or b) calculate it directly from sample data without using TTEST() function.

E.g.,

```p = TTEST(samples_1, samples_2, 2, 2)
```

is equivalent to -

```X1 = samples_1
X2 = samples_2
N1 = COUNT(X1)
N2 = COUNT(X2)
PV = (VAR(X1) * (N1 - 1) + VAR(X2) * (N2 - 1)) / (N1 + N2 - 2)
t = (SUM(X1) / N1 - SUM(X2) / N2) / SQRT(PV * (1 / N1 + 1 / N2))
p = TDIST( ABS(t), N1 + N2 - 2, 2)
```

*PV is pooled variance.

Hope this helps,

H

• Level 1 (0 points)

cool, I think I got it, thank you very much !

then, i'm using the TINV and the TTEST functions

thanks again!

• Level 5 (6,030 points)

You're quite welcome! Glad to be of help.

Best wishes,

Hiroto