3 Replies Latest reply on May 1, 2018 8:05 PM by swaroop.gantela

# How do you get this formula for calculating t-statistics to work

I'm trying to follow the formula above to calculate the t-statistics above for a two sample t test for comparing two means. This is what I tried but I got nothing in return.

Anybody who could help with me with this ?

• ###### 1. Re: How do you get this formula for calculating t-statistics to work

Dennis,

I don't think I got the details right, but I think the gist of it should work:

([Avg2]-[Avg1])

/

(SQRT([VarP1]/[Count1]+[VarP2]/[Count2]))

These subparts are further defined in the workbook attached

It looked like the values in Tableau were matching the rough

calculations I made in Excel on Sheet2 of the attached.

I wasn't sure what the delta was in the posted equation.

• ###### 2. Re: How do you get this formula for calculating t-statistics to work

Hi Swaroop,

If there are no trials, is there a way to estimate the variance using pooled variance ? What can you infer if I am just given the following ?

• ###### 3. Re: How do you get this formula for calculating t-statistics to work

Dennis,

These are deep waters.

Based on a cursory reading of

Pooled variance - Wikipedia

It looks like with a known pooled variance Sp, you could rearrange and get the ratio of S1 to S2.

In the t-stat equation, you appear to have all the parts except S1 and S2,

and from the above equation you could substitute S2 with (some constant)*S1.

But it still feels like you would need to know at least one of the actual variances in order

to get a final number.

Hopefully others in the Forum can chime in.

1 of 1 people found this helpful