Ecology project help

[jewel]

Topic -Effects of competition during larval stages on adult life history traits using butterflies

 

Hey everyone,

Thank you for having me (newbie here ). I was hoping if any one could guide us or help us (by 'us' i mean that i'm working in a group) as our project is based on the above mentioned topic. We've formulated the null and alternative hypothesis as;
null- there will be no effect of competition during larval stages on adult life history traits using butterflies

alternative -there will be effect of competition during larval stages on adult life history traits using butterflies

How can we design the experiments as we are using butterflies here , can anyone please help us to tell how to design effective experiments and all the things to take care while handling butterflies (we are planning to start with the order Ypthima a centrain kind of butterflies which are plenty here ),any help is welcome , like to give us the basics of an ecology experiments and things ,thank you

Edited August 10, 2016 by jewel

[Prometheus]

How do you plan to measure competition (binary?) and the life history traits?

 

So it sounds like you'll have one independent (competition) and several dependent (traits) variables? Is that right?

[jewel]

How do you plan to measure competition (binary?) and the life history traits?

 

So it sounds like you'll have one independent (competition) and several dependent (traits) variables? Is that right?

Yea there will be several dependent variables

[Prometheus]

So essentially you are testing multiple hypotheses: that competition effects trait 1, that competition effects trait 2,..., that competition effects trait n.

 

The first problem you have is that the type 1 error rate will be inflated. There are ways to mitigate against this; they generally involve larger sample sizes.(have you done a power calculation - did you know roughly what degree of difference in the traits to anticipate?).

 

Another way around this, especially if there are a very large number of traits, might be to consider dimensionality reduction techniques such as PCA or LASSO - particularly if you think the traits might have some roughly linear correlation. This will reduce the number of traits you need to consider, but may make interpreting results a little more difficult.

 

MANOVA, or similar, might also work, you'd only be testing one hypothesis - but then you will not be able to draw conclusions about individual traits, just the traits as a cluster.

 

Another approach, although i'm scraping the barrel here, would be to treat it as a classification problem. You'd then be answering the question: given certain traits, did the larva undergo competition. Would be handy if that is all you care about.

 

 

I'm just brainstorming though. You should actually consult a statistician - preferably one with ecology experience.

[jewel]

Thank you

 

So essentially you are testing multiple hypotheses: that competition effects trait 1, that competition effects trait 2,..., that competition effects trait n.

 

The first problem you have is that the type 1 error rate will be inflated. There are ways to mitigate against this; they generally involve larger sample sizes.(have you done a power calculation - did you know roughly what degree of difference in the traits to anticipate?).

 

Another way around this, especially if there are a very large number of traits, might be to consider dimensionality reduction techniques such as PCA or LASSO - particularly if you think the traits might have some roughly linear correlation. This will reduce the number of traits you need to consider, but may make interpreting results a little more difficult.

 

MANOVA, or similar, might also work, you'd only be testing one hypothesis - but then you will not be able to draw conclusions about individual traits, just the traits as a cluster.

 

Another approach, although i'm scraping the barrel here, would be to treat it as a classification problem. You'd then be answering the question: given certain traits, did the larva undergo competition. Would be handy if that is all you care about.

 

 

I'm just brainstorming though. You should actually consult a statistician - preferably one with ecology experience.

Thank you so much Mr. Prometheus , i'll search for it , and also can you recommend a good book for me, related to this

[Prometheus]

I don't know of one that covers all these subjects There should be some books on stats for ecology, that's your best bet, i know ecologists usually deal with wide data sets like yours.

[jimmydasaint]

What level of education are you jewel? K13? or A2 level ? That would influence the depth of your project and the depth of your responses.