Attached File(s)
-
2.JPG (47.7K)
Number of downloads: 22
Edited by Nrelo, 27 May 2009 - 03:13 AM.
0
7 replies to this topic
#1
-
- Active Members
-
- 78 posts
Enthusiast
Posted 27 May 2009 - 03:08 AM
I am analyzing the effect of a chemical at various concentration to different bacteria strains (see below). I really have no idea on which statistical test I should use. Should I use one-way ANOVA in this case?
#2
-
- Moderator
-
- 815 posts
Epigenetist
Posted 27 May 2009 - 06:16 AM
I agree. You should use one way ANOVA.
#3
-
- Active Members
-
- 206 posts
Veteran
Posted 27 May 2009 - 06:35 AM
Nrelo, on May 27 2009, 12:08 PM, said:
I am analyzing the effect of a chemical at various concentration to different bacteria strains (see below). I really have no idea on which statistical test I should use. Should I use one-way ANOVA in this case?
I'm not much of a stats guru, but to compare different treatments, and different strains... shouldn't you do 2-way anova?
Also, for 1-way anova I believe you'd have to introduce your data in a different way.
#4
-
- Active Members
-
- 83 posts
Enthusiast
Posted 28 May 2009 - 08:18 AM
Are those triplicates of each concentration??
I use to do that a lot with mammalian cells. Do you have PRISM?? its a wonderful software to analyze data like that.
I use to do that a lot with mammalian cells. Do you have PRISM?? its a wonderful software to analyze data like that.
#5
-
- Active Members
-
- 78 posts
Enthusiast
Posted 01 June 2009 - 01:05 AM
medchemgirl, on May 28 2009, 08:18 AM, said:
Are those triplicates of each concentration??
I use to do that a lot with mammalian cells. Do you have PRISM?? its a wonderful software to analyze data like that.
I use to do that a lot with mammalian cells. Do you have PRISM?? its a wonderful software to analyze data like that.
yes, they are triplicates. After I enter the data into prism, it shows a graph like below. But when I run one-way ANOVA, it compares the difference between group 2.5mM, 5mM and 10mM. But what I need to test is whether there are differences between strains, say between WT and M4 at 10mM concentration, how should I do that?
Attached File(s)
-
1.JPG (44.31K)
Number of downloads: 10
#6
-
- Active Members
-
- 19 posts
member
Posted 01 June 2009 - 08:51 AM
Anova is the test to use but only if you have confirmed the distribution of your data sets is Gausian (run a Kolmogorov Schmirnov or an Anderson Darling test though SPSS enables an add-on "homogeneity of variance" to the ANOVA). If the data isn't Gaussian, you could try a Kruskall Wallis test.
Use a one-way ANOVA when you can group the datasets only one way. I think you have two ways: by strain and by concentrations so go for a two-way ANOVA. Since you are using three concentrations, you actually have THREE ways of grouping your concentration data. You need to set this up carefully when you input the data.
ANOVA assumes the data is continuous and Gaussian. It then assesses whether any of the pairs of groups differ from the remaining group. It isn't so good at telling you which of the group-pairs differ. To elucidae that you need to do a post-hoc analysis. There are lots of such post-hoc tests available on most stats packages. Work with whatever your lab is familiar with.
Options could be Fishers's Test (aka Least Significant Difference test), Student Newman Keuls test (not the same as the Student's T-test which is for comparison of two groups in Gausian distribution) or a Tukey test.
If you need more, just shout.
Use a one-way ANOVA when you can group the datasets only one way. I think you have two ways: by strain and by concentrations so go for a two-way ANOVA. Since you are using three concentrations, you actually have THREE ways of grouping your concentration data. You need to set this up carefully when you input the data.
ANOVA assumes the data is continuous and Gaussian. It then assesses whether any of the pairs of groups differ from the remaining group. It isn't so good at telling you which of the group-pairs differ. To elucidae that you need to do a post-hoc analysis. There are lots of such post-hoc tests available on most stats packages. Work with whatever your lab is familiar with.
Options could be Fishers's Test (aka Least Significant Difference test), Student Newman Keuls test (not the same as the Student's T-test which is for comparison of two groups in Gausian distribution) or a Tukey test.
If you need more, just shout.
#7
-
- Active Members
-
- 78 posts
Enthusiast
Posted 02 June 2009 - 05:18 AM
Paraboxa, on Jun 1 2009, 08:51 AM, said:
Anova is the test to use but only if you have confirmed the distribution of your data sets is Gausian (run a Kolmogorov Schmirnov or an Anderson Darling test though SPSS enables an add-on "homogeneity of variance" to the ANOVA). If the data isn't Gaussian, you could try a Kruskall Wallis test.
Use a one-way ANOVA when you can group the datasets only one way. I think you have two ways: by strain and by concentrations so go for a two-way ANOVA. Since you are using three concentrations, you actually have THREE ways of grouping your concentration data. You need to set this up carefully when you input the data.
ANOVA assumes the data is continuous and Gaussian. It then assesses whether any of the pairs of groups differ from the remaining group. It isn't so good at telling you which of the group-pairs differ. To elucidae that you need to do a post-hoc analysis. There are lots of such post-hoc tests available on most stats packages. Work with whatever your lab is familiar with.
Options could be Fishers's Test (aka Least Significant Difference test), Student Newman Keuls test (not the same as the Student's T-test which is for comparison of two groups in Gausian distribution) or a Tukey test.
If you need more, just shout.
Use a one-way ANOVA when you can group the datasets only one way. I think you have two ways: by strain and by concentrations so go for a two-way ANOVA. Since you are using three concentrations, you actually have THREE ways of grouping your concentration data. You need to set this up carefully when you input the data.
ANOVA assumes the data is continuous and Gaussian. It then assesses whether any of the pairs of groups differ from the remaining group. It isn't so good at telling you which of the group-pairs differ. To elucidae that you need to do a post-hoc analysis. There are lots of such post-hoc tests available on most stats packages. Work with whatever your lab is familiar with.
Options could be Fishers's Test (aka Least Significant Difference test), Student Newman Keuls test (not the same as the Student's T-test which is for comparison of two groups in Gausian distribution) or a Tukey test.
If you need more, just shout.
thx a lot for the reply. I think I will try different ways of data input first
#8
-
- Active Members
-
- 78 posts
Enthusiast
Posted 04 June 2009 - 01:26 AM
hi again
I am just wondering whether I can run one-way anova separately. Say I compare the response of different strains at 2.5mM (same for 5mM and 10mM), so that the only variable is the strain. Then after all, if there are no differences, can I say "there are no significant differences between response of the strains at 2.5mM, 5mM and 10mM"?
I am just wondering whether I can run one-way anova separately. Say I compare the response of different strains at 2.5mM (same for 5mM and 10mM), so that the only variable is the strain. Then after all, if there are no differences, can I say "there are no significant differences between response of the strains at 2.5mM, 5mM and 10mM"?
