Which Statistical Test?

[MRaverz]

I'm trying to determine if any of four rates are significantly different from each other, but am unsure which statistical test to use.

Can anyone help me out?

 

[borrofburi]

What was the experiment? What was the mean and standard deviation of your samples for the four rates?

 

[lrkun]

I'm trying to determine if any of four rates are significantly different from each other, but am unsure which statistical test to use.

Can anyone help me out?

You need to give more information. Anyway, try Multivariate analysis of variance. It's simple and works well with two or more variables.

 

[MRaverz]

What was the experiment? What was the mean and standard deviation of your samples for the four rates?

Lemna growth rates.

24 replicates in 4 conditions.

I took the means of each of the 4 conditions and took the natural log of that data. I then put that on a graph of natural log of frond count over time so that I could find the estimated growth rate from the slope.

I'm now left with those 4 rates and am unsure how to work out if they're significantly different from each other.

The rates are 0.7, 0.3, 0.5 and 0.6.
Standard deviation is 0.17078, I believe and variance is 0.0292.


I have a feeling I'll be needing to do something with the change in frond numbers. D:

 

[MRaverz]

I'm trying to determine if any of four rates are significantly different from each other, but am unsure which statistical test to use.

Can anyone help me out?

You need to give more information. Anyway, try Multivariate analysis of variance. It's simple and works well with two or more variables.

I think it might be a case of using univariate ANOVA, seeing as that came up in one of our stats lectures.

I fear doing the wrong one though. D:

 

[lrkun]

I think it might be a case of using univariate ANOVA, seeing as that came up in one of our stats lectures.

I fear doing the wrong one though. D:

But a univariate / one way analysis of variance works only with a single variable right? Y.Y

I think I need to review my statistics. HEhe

 

[MRaverz]

I think it might be a case of using univariate ANOVA, seeing as that came up in one of our stats lectures.

I fear doing the wrong one though. D:

But a univariate / one way analysis of variance works only with a single variable right? Y.Y

I think I need to review my statistics. HEhe

Isn't the single variable the independent variable? That being initial population number. The rest was controlled for.

And I think I need to learn my stats.

 

[lrkun]

Isn't the single variable the independent variable? That being initial population number. The rest was controlled for.

And I think I need to learn my stats.

What's your independent variable in this case?

 

[MRaverz]

Isn't the single variable the independent variable? That being initial population number. The rest was controlled for.

And I think I need to learn my stats.

What's your independent variable in this case?

For the first, initial number of fronds.
For the second, concentration of nutrients.

 

[lrkun]

For the first, initial number of fronds.
For the second, concentration of nutrients.

Will you be conducting two different one way tests? or will you do one with both factors?

 

[MRaverz]

For the first, initial number of fronds.
For the second, concentration of nutrients.

Will you be conducting two different one way tests? or will you do one with both factors?

Two one-way ANOVA tests, the factors are separate in this case.

I'm guessing that if I were to combine them, I'd need a two-way ANOVA?

 

[lrkun]

Two one-way ANOVA tests, the factors are separate in this case.

I'm guessing that if I were to combine them, I'd need a two-way ANOVA?

^-^ yes.

 

[MRaverz]

Two one-way ANOVA tests, the factors are separate in this case.

I'm guessing that if I were to combine them, I'd need a two-way ANOVA?

^-^ yes.

Thanks!

 

[borrofburi]

What was the experiment? What was the mean and standard deviation of your samples for the four rates?

Lemna growth rates.

24 replicates in 4 conditions.

I took the means of each of the 4 conditions and took the natural log of that data. I then put that on a graph of natural log of frond count over time so that I could find the estimated growth rate from the slope.

I'm now left with those 4 rates and am unsure how to work out if they're significantly different from each other.

The rates are 0.7, 0.3, 0.5 and 0.6.
Standard deviation is 0.17078, I believe and variance is 0.0292.


I have a feeling I'll be needing to do something with the change in frond numbers. D:

In that case: I have no clue. I've never been good at statistics. I just knew that this kind of background info was, probably, necessary to answer your question.

 

[Master_Ghost_Knight]

I am not a biologist and so i am not aware of the full dinamics of the problem. But from a strict closed box prespective, it does appear that a difference of 0.3 to 0.7 (i.e. more than 100% difference) should be significant, but then again I do not know what 0.3 or 0.7 measures, is it a ratio or an absolute value? Given that I can't really say. To tell if it is significant or not you need to account for various factors, such a precision of the measuring equipments and models or the iner variance of the population within the same conditions and then you would need. If you have time i would advise some research into metriology.

 

[MRaverz]

I am not a biologist and so i am not aware of the full dinamics of the problem. But from a strict closed box prespective, it does appear that a difference of 0.3 to 0.7 (i.e. more than 100% difference) should be significant, but then again I do not know what 0.3 or 0.7 measures, is it a ratio or an absolute value? Given that I can't really say. To tell if it is significant or not you need to account for various factors, such a precision of the measuring equipments and models or the iner variance of the population within the same conditions and then you would need. If you have time i would advise some research into metriology.

You need not be a biologist, but a statistician.

Plus those values were incorrect, I've corrected them now.

As such the test I ended up using compared the variance within a group of rates to other groups of rates. The first seemed to not be statistically significantly different, the second was significant but in the wrong direction.

Therefore, according to the data I collected, the less resources you have, the greater the rate of growth. Something tells me the practical went wrong.

 

[Master_Ghost_Knight]

You need not be a biologist, but a statistician.

Well, from my experience pure statistician are awful statisticians. Because a variance of 0.2 maybe significant if the natural variance of the rate is very strict, or maybe it isn't if you could get about that whitout chaging any variables. But this isn't exactly blind, you have to know what is the variance you already would expect just by doing nothing and how good is your measurment. Just looking at a couple of numbers withtout any context and be able to tell if they are significant or not is just not something we should do, because significance in the latin sense is very subjective and you must have an objective criteria to determine "significance" in science. Yes we can do it objectively, but you must account for a variety of factors that I am simply unaware of in your particular problem. People do need to get trained and do need to know some background in order to be able to look at data and make say "Oh this is interesting and meaningfull" or "this is just trivial and uninteresting". Sure you can give the data to a statitician and let him/her come up with a number, but you need another meaningfull number to compare it to and be able to say "this works" or "this doesn't".

Ps. Kind of the all point of this topic since you are having problems in telling if the growth rate is significant or not.