The Student Room Group

Chi Squared Test

Hi, I need to carry out a statistical test. The experiment I did was the effect of colour of light on seedling growth, the data is cm. I carried out the chi squared test but it came out as 50% probability but there is a significant difference between red/blue & green/yellow. I asked my tutor if it would work for more than 2 groups and he said it would but I am not sure ?? I attached my data so you can get a better idea of what i'm talking about.

please help :smile:
Original post by IchHabDichLieb
Hi, I need to carry out a statistical test. The experiment I did was the effect of colour of light on seedling growth, the data is cm. I carried out the chi squared test but it came out as 50% probability but there is a significant difference between red/blue & green/yellow. I asked my tutor if it would work for more than 2 groups and he said it would but I am not sure ?? I attached my data so you can get a better idea of what i'm talking about.

please help :smile:



What was your alpha and what was your degrees of freedom (v)?
(edited 10 years ago)
Reply 2
Original post by ExWunderkind
What was your alpha and what was your degrees of freedom (v)?


my alpha should be 0.05 but it's 0.5

from what i've read about degree of freedom on the internet I think it's 4 as I have 5 groups, right?

this is a biology experiment by the way so I don't have statistical knowledge
(edited 10 years ago)
Original post by IchHabDichLieb
what is alpha?

from what i've read about degree of freedom on the internet I think it's 4 as I have 5 groups, right?

this is a biology experiment by the way so I don't have statistical knowledge


Alpha is the significance that you are testing at, or the area shaded on your graph.

Degrees of freedom is (number of rows - 1) * (number of columns - 1).
Reply 4
it should be 0.05 but it's 0.5.

which rows and columns? on my raw data or the chi squared table?
Original post by IchHabDichLieb
my alpha should be 0.05 but it's 0.5

from what i've read about degree of freedom on the internet I think it's 4 as I have 5 groups, right?

this is a biology experiment by the way so I don't have statistical knowledge


Sorry I just caught your edit.

You set the alpha, or significance you are looking at. For example if you are testing at a 0.05 significance level, you find that alpha and read down the table to your degrees of freedom.

Then look to see if your Chi^2 is greater or lesser and whether you accept your null or alternative hypothesis.
Original post by IchHabDichLieb
it should be 0.05 but it's 0.5.

which rows and columns? on my raw data or the chi squared table?


The rows and columns for working out your V is your data table.
Reply 7
Original post by ExWunderkind
The rows and columns for working out your V is your data table.


raw data table?
i have 32 rows and 14 columns on my raw data table
Original post by IchHabDichLieb
raw data table?
i have 32 rows and 14 columns on my raw data table


Ok I've just looked at your table and I'm not sure that a Chi Squared test is suitable for your experiment.

For it to be used your data must have the following characteristics:

1) The data must be in the form of frequencies.

2) The frequency data must have a precise numerical value and must be organised into categories or groups.

3) The expected frequency in any one cell of the data must be greater than 5.

4) The total number of observations must be greater than 20.

Does your data fit this?
Reply 9
Original post by ExWunderkind
Ok I've just looked at your table and I'm not sure that a Chi Squared test is suitable for your experiment.

For it to be used your data must have the following characteristics:

1) The data must be in the form of frequencies.

2) The frequency data must have a precise numerical value and must be organised into categories or groups.

3) The expected frequency in any one cell of the data must be greater than 5.

4) The total number of observations must be greater than 20.

Does your data fit this?


my tutor said both chi squared and man whitney u test would be suitable.
is there any statistical test that would be suitable?
Original post by IchHabDichLieb
my tutor said both chi squared and man whitney u test would be suitable.
is there any statistical test that would be suitable?


I think it would be suitable but you would have to arrange your raw data into categories. You are measuring seedling growth in cm right?

For that to work you would need your data tabulated keeping the current columns as they are, but your rows would be something like 0< cm >1 and 1< cm >2 and then you count them as a frequency.

If you are unsure, is it possible for you to show me your raw data?
Original post by ExWunderkind
I think it would be suitable but you would have to arrange your raw data into categories. You are measuring seedling growth in cm right?

For that to work you would need your data tabulated keeping the current columns as they are, but your rows would be something like 0< cm >1 and 1< cm >2 and then you count them as a frequency.

If you are unsure, is it possible for you to show me your raw data?


it's on sheet 1
Original post by IchHabDichLieb
it's on sheet 1


Sorry I've taken so long to reply, but been really busy.

I've looked at your original data set and I'm afraid it isn't useable for Chi^2.

I've attached my workings for you to look at and view, and you can see where there are Expected Values of 0, this really throws things out.

You might need to look at the other statistical test your tutor mentioned, but see what you think.
Original post by ExWunderkind
Sorry I've taken so long to reply, but been really busy.

I've looked at your original data set and I'm afraid it isn't useable for Chi^2.

I've attached my workings for you to look at and view, and you can see where there are Expected Values of 0, this really throws things out.

You might need to look at the other statistical test your tutor mentioned, but see what you think.


thanks for your effort. would it be suitable if I just used two groups? say red and green?
Original post by IchHabDichLieb
thanks for your effort. would it be suitable if I just used two groups? say red and green?


That isn't the problem, the test is suitable for the amount of categories you have.

What the problem is, is that you don't have high enough expected values. If any of the expected values are below 5 then this test is unsuitable.
what is chi?
Original post by TeddyBasherz
what is chi?


It's a greek letter. Pronounced "kai".
Original post by ExWunderkind
It's a greek letter. Pronounced "kai".


what if I changed it to millimeters instead of cm?

Quick Reply