Not sure what Statistical test to use......

seanknowsall

I am doing a project on a particular habitat and have two areas that I want to compare. Area A has been grazed by sheep and Area B has been grazed by cattle. The two types of grazing has had a different impact on the populations of wildflowers growing here. I have to represent this using some type of stats table... before grazing commenced I did a survey of the species and recorded 51....all 51 species were common to both areas. The area was then divided and grazing started for a set period of time. A year later I went back and found that Area A (sheep) had only 24 of the original 51 whereas Area B (cattle) had 40. My hypothesis is that cattle grazing is a more wildflower friendly regime on unimproved grasslands. I doing an ecology masters and so far have managed to get away without knowing anything about stats...until now!! Any suggestions gratefully accepted!!

Ostrom

seanknowsall wrote: »

I am doing a project on a particular habitat and have two areas that I want to compare. Area A has been grazed by sheep and Area B has been grazed by cattle. The two types of grazing has had a different impact on the populations of wildflowers growing here. I have to represent this using some type of stats table... before grazing commenced I did a survey of the species and recorded 51....all 51 species were common to both areas. The area was then divided and grazing started for a set period of time. A year later I went back and found that Area A (sheep) had only 24 of the original 51 whereas Area B (cattle) had 40. My hypothesis is that cattle grazing is a more wildflower friendly regime on unimproved grasslands. I doing an ecology masters and so far have managed to get away without knowing anything about stats...until now!! Any suggestions gratefully accepted!!

I'm not sure you need a formal test - did you measure only once (before and after) for each group? Usually a basic between-groups test of significance would involve running an independent samples t-test, but your data are not sample data, and all the t-test will tell you is the probability of observed difference due to chance.

Do you have any repeat measures over time, or just these two data points? If so, then there may be a few further options. Otherwise, (and I'm open to correction here), I think you might be limited to descriptives.

seanknowsall

efla wrote: »

I'm not sure you need a formal test - did you measure only once (before and after) for each group? Usually a basic between-groups test of significance would involve running an independent samples t-test, but your data are not sample data, and all the t-test will tell you is the probability of observed difference due to chance.

Do you have any repeat measures over time, or just these two data points? If so, then there may be a few further options. Otherwise, (and I'm open to correction here), I think you might be limited to descriptives.

Thanks for your post. I only did a plant species population count at start of the test and then a year later after grazing had finished. This is a mock paper I writing... the data doesnt have to be real it is in fact made up... I wanted to try to make it look more journal like by throwing in some stats. Do you have any ideas what I could add in to do this? I can just use a descriptive analyisis I guess if my data and samples are to simple. Thanks

Ostrom

seanknowsall wrote: »

Thanks for your post. I only did a plant species population count at start of the test and then a year later after grazing had finished. This is a mock paper I writing... the data doesnt have to be real it is in fact made up... I wanted to try to make it look more journal like by throwing in some stats. Do you have any ideas what I could add in to do this? I can just use a descriptive analyisis I guess if my data and samples are to simple. Thanks

If I'm understanding you, this is some kind of thought exercise for an upcoming project? If you are comparing only these two areas, then statistical inference is not appropriate - such tests merely tell you whether your observed results are likely to be 'real' or chance effects, and they require sample data (usually with a minimum sample size).

In order for such a test to be appropriate, you would need multiple sites of each type (i.e. 20 sheep grazed and 20 cattle grazed areas). In this case, you could compare average species levels across both groups at t2 - but here you may need slightly more complex tests, given that you have different interventions / different initial states - and if you manipulate or observe other variables which may contrbute to the different counts, you will then need some kind of multivariate technique.

If your study is designed to include measures at multiple time points, you're opening up the possibility of more complex models and tests.

If you are trying to determine an appropriate sample size, it might be helpful to start with a statistical test selection sheet - have you taken any stats course before? These are step diagrams where you select a final test based on the objectives of your study (i.e. how many groups do you want to compare, what kind of predictors you are using).

You could subsequently use a statistical power or sample size calculator to work out a minimum suitable sample size. From the sounds of it, you probably dont need anything too complicated, and a lot of this stuff is deceptively simple. A good supervisor will be able to point you in the right direction in minutes, as long as you are clear about your research design.