retired_nashfine

Thanks will have a look just now :thup:

retired_mountainsong

maths like this honestly like another language to me, anybody who can do this stuff has props from me, I am awful at maths

retired_nashfine

Thats great Andy the only question I have now is how do you create the critical Chi?

retired_kimmymarlo

frankly, I just know how to do it on Excel, which is =CHIINV(probability, degrees of freedom).

I used an alpha of 0.05, as that's convention, and 4 degrees of freedom, as (number of rows - 1) times (number of columns - 1) = (5-1)(2-1) = 4.

retired_nashfine

Kind of figured it out, based on degree's of freedom and the amount of questions right?

Therefore the 0.05 used in the spreadsheet is the same as the 5 questions used.

Thus it calculated the critical.

So to calculate the same for other questions I would only change the degrees of freedom to whatever stated within SPSS.

That right?

If so thats another hypothesis done

retired_nashfine

Was slightly out with the 0.05, am I right in thinking it would be a constant for the questions?

Pretty sure SPSS is giving me exactly what you are stateing its just good to understand the background of it.

retired_kimmymarlo

no, degrees of freedom are a function of how many variables you have in your table.

if there were 4 groups and 4 conditions, you would have (4-1)(4-1), or 9, degrees of freedom. alpha corresponds to the degree of uncertainty you're willing to accept, as 1 - alpha (or 1 - 0.05 = 0.95) generally means the probability that you're actually right. the smaller your alpha, the most likely you're correct, but the less predictive power your test has, given that it widens the amount of "correctness" that is possible, so to speak.

retired_kimmymarlo

yeah, everything you need to know is there. the "0.010" in the first line corresponds to the probability from the chi-square distribution with a test stat for 13.301 and 4 df; as it's smaller than alpha, .05, you reject the null.

retired_nashfine

ahhhhhhhh

But if I am trying to prove a hypothesis over a range of questions I am best off keeping alpha the same. To allow for good comparison...

Think I am to grips with it now

retired_kimmymarlo

yep.

any more questions, up this thread.