# Thread: 4 groups test - is it possible?

1. ## 4 groups test - is it possible?

My problem -

I have conducted an experiment which grows Water Cress in 6 different
growing mediums (sand, gravel, stone, peat, compost, fertilized
compost).

The aim is to see which growing medium is most effective by measuring
the stems in (mm) at a sample of 50 stems per pot/medium

In 2 of the mediums, the water cress did not germinate and therefore I
have no data recorded. I have recorded raw data for the other 4
growing mediums which totals 200 measurements, have taken averages for
each and drew an ogive and bar chart to show this.

The problem is that I don't know which hypothesis test to use to show
my Inferential statistics. I think my data is Interval as im
measuring using a ruler? and also think that Im using an Independent
measures design? which leads me to believe I need to use a Mann
Whitney 'U' test.

next week.

I reg'd under a diff name but I never received the email to activate.

Anyway,
Thanks in hope.

2. Use a 1-way ANOVA (Analysis of Variance). This will show if any of the 4 mediums are different from the rest, and then you can follow up this test with post-hoc comparisons (similar to t-tests) if necessary.

3. I am very grateful for your reply John. I shall look into this ANOVA test as I've never heard of it before (I hope it's easy)

My head is sore which all the stress over this

4. I just looked the example and am now going to attempt it. It looks quite handy if I can keep my concentration.

5. ANOVAs can take a little work, that's for sure. However, if you have Microsoft Excel, then you're in luck.

Here's a quick tutorial on how to calculate ANOVAs using Excel:
http://www.isixsigma.com/library/content/c021111a.asp

Of course, it's even easier with actual stats software...

6. Yeah did it with excel. Just needed to manipulate my table a little to accomodate the program.

F P-value F crit
762.2156302 6.4909E-137 2.408491184

I think it's safe to say I can reject the null hyp in this case? lol

7. H1 = there will be a diff of growth within diff mediums etc etc

H0 = there will no be diff.....

8. Would it be logic to reject my null hypothesis when the F - 762.22 is greater than the F crit - 2.41

Thoughts?

9. Yes. But more useful to your research would be to plot the group means and their confidence intervals, or at least show them in a table. The initial F test just tells you whether a difference exists - you'll need to dig a little further to see exactly what it is.

10. Yep I have drawn a bar graph and also an Ogive to show the group means. Looks kinda cool as I have used the textures of the various mediums for each bar!

Am looking good scores for this.

Guy I really appreciate your input. I owe you a drink for sure.

11. The attached image in the ANOVA test example isn't working for me.

12. My p value has came out at -

P-value
6.49

Wouldn't that mean that I would accept the Null hypothesis?

13. The image file isn't there for some reason - don't know. Re-check your p-value (it's a probability, so it can only take on a value between 0 and 1)

14. This is what excel has gave me -

ANOVA
Source of Variation-----SS------df------MS---------F----------P-value
Between Groups 502077.72-----4-----125519.43---762.22-----6.4909E-137
F Crit = 2.41

Within Groups----40345.88-----245----164.68

Totals-----------542423.6-----249

15. it's giving you the p-value in scientific notation

6.4909E-137 = 6.4909 * 10^(-137) --> an extremely small number - you need to take 6.4909 and move the decimal point 137 places to the left, which is for all practical purposes = 0, so reject Ho.