Help with data analysis please - due Thursday

[shadowthecat]

I am finished my report but I am not confident about my conclusion. I am hoping to find some statistical calculations that I can do in excel to verify that the changes in my data aren’t significant but are just random.

I tested the growth of radishes. I did 4 trials, 6 seeds each trial, which is 24 seeds in total for each group. I had 3 different groups. The data in the 3 groups is very similar when graphed but the control group is slightly larger on average but I don’t think that it is because of the independent variable. I think that this difference is because random seed differences, but I am not sure.

I spent a lot of time searching the web and science buddies posts about statistics but me and my parents are confused. When I calculate standard deviation the difference within each group is quite a bit larger than the deviation between the groups. Does this mean that the groups are pretty much the same or is there another type of calculation that I can use? Is a t-test a proper calculation? This is a short summary of my data

Root Control Shoot Control Root Group 1 Shoot Group 1 Root Group 2 Shoot Group 2
Count 24 24 24 24 24 24
Average 9.71 5.2 8.44 5.3 9.07 5.06
Standard
Deviation 2.44 1.55 2.34 1.48 3.69 1.58

[SciB]

Hi shadowthecat,

The way to compare two means to determine if they are the same or different is to use the Student's t test (http://en.wikipedia.org/wiki/Student's_t-test). With your data you would be comparing the root control mean value to the root 1 mean value then the root control to the root 2 mean. Then do the same thing for the shoot control and 1 and 2 means.

You would be using a 2-tailed t-test for unpaired means. You can do this in Excel by inputting your data and selecting t test as your function. The program will do the calculation for you and give you the comparison value. Remember that you are testing the null hypothesis which says that your control and test means are NOT different. So if the t-test supports the null hypothesis then you can conclude that there is no difference.

Try the t-test and see what you get. If you need help again, send another query. Don't worry! We'll help you get your calculations done before Thursday.

Thanks for using ask-an-expert.

SciB

[shadowthecat]

Thank you VERY much for your help. I would like to check if I’ve got this right.

I calculated TTEST in excel using 2 sets of data. I entered 2 in the formula for 2-tail and 2 again for the type because they are unpaired & I assume that I have equal variances. My Formula looked like this =TTEST(array 1,array2,2,2) This is my t scores

control VS group 1 root = 0.77505093
control VS Group 1 shoot = .823659717
Control VS group 2 root = .487131027
Control VS group 2 shoot = .766860071

I compare the t score to the table of critical value for t when two-tail alpha=.05.
Degrees of freedom =n-1 which is 23 for my sample of 24 seeds
The table shows that the critical value of t =2.0686 .
If my scores are less than 2.0686 then I can accept the null hypothesis and say that the means of each group are equal to each other with 95% confidence (because alpha is .05)????

I already calculated standard deviation before. Should I still mention that the deviation within is group is much higher than the deviation between groups? Or is this not important?

Thank you

[SciB]

Hi shadowthecat,

You did a great job using the t-test, but there's one correction I think you need to make, and that's on the degrees of freedom. When you are comparing TWO groups the formula for degrees of freedom is DF = (n1 + n2) - 2, not n - 1. This is explained in the Handbook of Biostatistics here: http://udel.edu/~mcdonald/statttest.html

Recalculating the DF will not change your outcome. The null hypothesis is still correct and there's no difference between the controls and the experimentals.

Good job--and you made it before Thursday!

Cheers,

SciB

[shadowthecat]

Thank you again.

can I just confirm that it is correct to say that confidence is 95%? Is this what it means when alpha is .05?

[SciB]

Yes. The 0.05 alpha value means that there's only a 5% probability that you are wrong in accepting the null hypothesis---95% that you are correct.

Some people use an alpha of 0.01 which is equivalent to 99% confidence, but that is usually not necessary.

Congratulations! Please use the expert forum again for your next project.

Cheers,

SciB