# Thread: Singificance levels, critical values and p-value. Help!!!!

1. ## Singificance levels, critical values and p-value. Help!!!!

Dear Maths wizards,

Firstly, i hope my post is in the correct place.

I am not a mathematician but for a paper i am writing on music and education, i have conducted a study to find out if music capital affects pupils' engagement in classroom music lessons.

I am currently analysing my data which is what brought me to your forum. I have two variables: a pupil's level of Musical Capital [MC] and a pupil's level of engagement in lessons. Both of these variables have been coded into numeric form and i am looking at their relationship to see if MC affects engagement. I have so far done quite a lot of statistics research and i understand a lot more than i thought i would, but need a couple of things clarifying...

Would you agree that with the data i have and the fact that im looking at a relationship between two variables, i can use a null-hypothesis, a significance level and a p-value? All these look good to me, but i feel like im asking the 'chicken and the egg' question. Which of these comes first? Would you write about a null-hypothesis in your findings of a study or before? Do you decide on a 0.01 or 0.05 depending on how reliable you think your hypothesis is or is there a formula you have to apply to it? Once you have got this, how do you find your p-value (and what exactly is a p-value?!)

As i said, im sorry if i have posted this in the wrong thread...

Any help on the questions i have asked as well as new advice would be so welcome.

Yours confusedly,

Olivia

2. I'd use regression on your two variables.
Based on your statement 'to see if MC affects engagement'
I would like to see if there is a positive slope when fitting the data.
I'd use 'MC' as X and 'engagement' as Y.
Testing to see if the slope is positive is the same as testing to see if there is a positive correlation.
And we can use all kinds of models too, but I would start with $y=\beta_0+\beta_1 x+\epsilon$.
I would just obtain the p-values of your tests and I might test for normality before I run those t tests.
I don't think I'm a wizard, I never even played dungeons and dragons as a child.

3. Thank you for your advice! Yes, i have set the two variables on an X and Y axis and they present a positive correlation (0.74 correlation coefficient).

As for the p-value, t-tests and testing for normality...

the p-value i am using is 0.05 (95% sure the results could not be found by chance- is this correct?)

i have an f-test value of 9.98904E-16, which i think means my variances are not equal, therefore i have to use un-equal variance in the t-test. (Am i on the right tracks?)

I am not sure whether to use a one tailed or two tailed t-test. I do know the supposed direction of difference of my findings, but am i right in thinking its just safer to use a two tailed test in all cases?

I have used excel to find out the t-test value which is 0.001117649. In fact, it is the same for a one-tailed and two-tailed test. What does this show me?

Now that i have my t-test figure, what does it mean? Do you know of a good stats website that will tell me what this number is showing? Its all very well having it, but i would like to know how to use it!

What did you mean by the 'test for normality'? Do you mean the p-value?

Thank you again, and please yell if i'm on the wrong lines!

4. I have no idea what your hypothesis is when you say that you you 'have an f-test value of 9.98904E-16'.
You need to tell me what the hypotheses are and what you are trying to test.
And I don't think you understand what a p-value is.
It's the probability of being worse off than your test statistic.
So I need the alternative hypothesis and the test stat to determine the p-value.
But I also need to know what kind of test you're running.