chi squared test and null hypothesis

In my book it says that the chi squared test tests the null hypothesis which is that 'there's no significant different between the observed and expected numbers, and any difference is due to chance'.

But I thought the null hypothesis of an experiment would be: like if we were testing the link between obesity and heart disease then it would be 'there would be no relationship between obesity and heart disease'.

I'm confused...

Reply 1

Gnome :)

The book is right. Using your example, the null hypothesis would be "There is no differrence beteen the incidence of heart disease in obese people compared to what we'd normally expect" or something along those lines.

Reply 2

username895805

Original post by Gnome :)

Ok, it's just in psychology in our experiments that's how the null hypothesis would be written, or maybe it's just like that with chi squared tests, what I mean is- the chi squared test has it's own experimental and null hypothesis? The null being the any difference between the observed and expected values is due to chance

Reply 3

Gnome :)

Original post by celina10

For biology:
-Chi squared: "there is no significant difference between observed and expected values"
-Spearman's Rank: "there is no relationship/correlation between X and Y"
-Standard error and 95% confidence limits: "there is no significant difference between X and Y"

Reply 4

pjm600

Same as above for geography. The null hypothesis states that there is no correlation between data sets.

Reply 5

username895805

Original post by Gnome :)

Original post by pjm600

Same as above for geography. The null hypothesis states that there is no correlation between data sets.

So it's just with statistical tests where the null hypothesis may be different, depending on what the test is trying to find out?

By the way, why is the critical value for the chi square test so low? I mean our results would be acceptable if there's a 5% probability that the difference in our results is due to chance :confused:

, isn't that very inaccurate as that would mean that there's a 95% chance that our value was not by chance and that the theory is wrong.

(edited 10 years ago)

Reply 6

Gnome :)

Original post by celina10

, isn't that very inaccurate as that would mean that there's a 95% chance that our value was not by chance and that the theory is wrong.

Err, using the degree of freedom for p=0.95 means that there is a 95% probability that the differences are due to a biologial factor and not due to chance.

Reply 7

Anoth3rhobo

Ignore the last comment, I've realised I've made some mistakes.

We're all human :smile:

(edited 10 years ago)

Reply 8

username895805

Original post by Gnome :)

Err, using the degree of freedom for p=0.95 means that there is a 95% probability that the differences are due to a biologial factor and not due to chance.

But in my book the cut off point is 0.05:

It's not that clear so I'll quote it here: 'critical value of X^2 0.05 p level; this is the level at which we are 95% certain the result is not due to chance, agreed on by statisticians as a cut off point'

The green part says 'accept null hypothesis (any difference is due to chance and not significant)'

The red part says 'reject null hypothesis; accept experimental hypothesis (difference is significant, not due to chance)'

(edited 10 years ago)

Reply 9

Gnome :)

My mistake, meant p=0.05.

Remember, using chi squared you are looking for a significant difference. Therefore a 5% probability that results are due to chance means you are 95% sure it's due to a biological factor. As a researcher, you want to reject the null hypothesis, as this shows there is a significant difference.

Reply 10

username895805

Original post by Gnome :)

I understand what you're trying to say and it makes sense, but the book says something different. Basically before the table it talks about how meiosis causes a different combination of alleles to be present and the expected phenotypic ratio we'd expect is 9 : 3: 3 : 1. But if the observed value is different we apparently have to use this equation to see if the difference was due to chance and isn't significant, but if the difference is significant i.e. the percentage that the results were due to chance alone was below 5% then we have to re-think our current biological theory. But what I'm confused about is: if we're 5% certain that the deviation from the expected value is due to chance alone, then that would mean that we're 95% certain that our theory is wrong :confused:

anyone?

Original post by celina10

I really don't know how to explain it any other way.

So, you set a null hypothesis, do your experiment, and use the formula to get a value for X^2. You then compare this value to the relevant value in a table of degrees of freedom. If your X^2 value is less than the one in the table, there is a 95% probability that any differences are due to chance, so you accept the null hypothesis (no significant difference). If your value is higher than the one in the table, there is a 95% probability that differences are due to a biological factor, so you reject the null hypothesis (significant difference).

I think you're confusing yourself; the book is unlikely to be wrong on such a basic component of the course.