# Chi-squared analysis.

• Sep 22nd 2010, 09:50 PM
Lucy1
Chi-squared analysis.
Ok so I am thoroughly confused.

I am measuring the sex ratios of Drosophila melanogaster, and I am having issues interpreting my data for my report.
My report is looking at whether increasing concentrations of estrogen will produce more male flies.

If my chi-squared number is larger than 3.481 for df =1. But my p-value>>0.05.

How do I interpret this?

Also what if both my chi-squared value are larger and p-value>0.05 also.

And finally if chi-squared value lower than 3.481 but p-value still higher

I will be so grateful to anyone who can explain this to me!
• Sep 22nd 2010, 10:42 PM
pickslides
When testing at 0.05 you will have a chi-squared 'critical value' that corresponds to this significance.

If you’re chi-squared calculated value is less than this critical value then p will always be > 0.05 and you do not reject the null hypothesis.

If you’re chi-squared calculated value is greater than this critical value then p will always be < 0.05 and you can conclude there is evidence to reject the null hypothesis.
• Sep 22nd 2010, 11:46 PM
Lucy1
How can p always be something when i get these values from R ( program for stats.)
The critical value for df=1 @ p = 0.05 = 3.84.

If these are the values i have gotten , how can my p value be below 0.05? It's clearly not.
And just to check , if i am checking to see wether the estrogen influences the sex to shift toward more females which would be my null hypothesis?

null= Estrgoen has no effect on sex determination.
Alt= Estrogen does have an effect on sex determination.

Is that right , also below are my values.

Control Vs 0.1 (0.0075µM)
X-squared = 0.125, df = 1, p-value = 0.7237

Control vs 1% ( 0.075 µM)
X-squared = 1.2564, df = 1, p-value = 0.2623

Control Vs 10% (0.75 µM)
X-squared = 1.3243, df = 1, p-value = 0.2498

For 0.1 Vs 1%
X-squared = 0.6052, df = 1, p-value = 0.4366
• Sep 23rd 2010, 12:29 AM
Lucy1
ok so i just spend like an hour trying to get this and figured i'd just read it wrong , I get what you mean. Thank you for your help!!!
• Sep 23rd 2010, 02:21 AM
CaptainBlack
Quote:

Originally Posted by Lucy1
The critical value for df=1 @ p = 0.05 = 3.84.

Don't do this it is confusing what you mean is:

The critical value with one degree of freedom at p=0.05 is 3.84.

CB
• Sep 23rd 2010, 02:26 AM
Lucy1
ok so just to clarify.

Re: the data i posted above.

Null : Increasing concentrations of estrogen does not increase female to male ratio.
Alt : Increasing concentrations of estrogen increases female to male ratio.

Using data posted in previous post, all result in Chi- squared > 3.841 , and p>0.05 so there is no evidence to reject the null hypothesis.

Is this correct?
• Sep 26th 2010, 01:54 PM
pickslides
Quote:

Originally Posted by Lucy1
Null : Increasing concentrations of estrogen does not increase female to male ratio.
Alt : Increasing concentrations of estrogen increases female to male ratio.

Sounds good.

Quote:

Originally Posted by Lucy1

Using data posted in previous post, all result in Chi- squared > 3.841 ,

$\displaystyle \displaystyle \chi ^2 = 3.841$

Quote:

Originally Posted by Lucy1
and p>0.05 so there is no evidence to reject the null hypothesis.

correct
• Sep 29th 2010, 09:35 AM
Pookabear
Not wanting to confuse the issue at all, but in my uni course we've always been told that H0 for chi-squared is that there is no association, and that H1 is that there is some association, but you don't know what the association is until you do the goodness of fit testing.
Are they teaching us the 'easy' way, and there actually are different hypotheses to test for chi-squared?
• Sep 29th 2010, 09:36 AM
Pookabear
Not wanting to confuse the issue at all, but in my uni course we've always been told that H0 for chi-squared is that there is no association, and that H1 is that there is some association, but you don't know what the association is until you do the goodness of fit testing.
Are they teaching us the 'easy' way, and there actually are different hypotheses to test for chi-squared?
• Sep 29th 2010, 02:07 PM
pickslides
In general I would say

$\displaystyle H_0:$ Observed = Expected.

$\displaystyle H_a:$ Observed differs from Expected.

This is not the only $\displaystyle \chi^2$ test as such. $\displaystyle \chi^2$ is a distribution that applies to many tests.