# Thread: Statistical analysis, this is impossible

1. ## Statistical analysis, this is impossible

Firstly, thank you to anybody who is willing to help. No math teacher can help me here so i'm relying on your help.

1. I conducted a series experiments to measure CO2 given off of yeast in a 30 minute window at different temperatures, of course there was a 'normal' correlation displayed (less activity as we stray further from optimum enzyme temperature):

Null Hypothesis:- There will be no difference in the rate of respiration as the temperature is manipulated.
Experimental Hypothesis:- There will be a difference in the rate of respiration as the temperature is manipulated.

2. I then plotted each individual experiment (refer to image) as a scatter plot, this displayed curvilinear? data as the CO2 production slowed down towards the end of the experiment

3. I calculated the gradient of the graph (Delta Y/Delta X), giving me the rate of reaction for each experiment.

4. I need to conduct statistical analysis into this data. I tried using chi-squared however the p value stated the results were down to chance; obviously rubbish

Am I using the test improperly or should I be using a different test? I think it is interval data, please suggest a test I could possibly use, bearing in mind the rate values follow a normal correlation in relation to temperature (peaking at 50 degrees)

I have included pictures of the scatter plot, also of the rate values in relation to temperature and of course the attempt of a chi-squared test.

2. ## Re: Statistical analysis, this is impossible

Hey pwillia.

I took a look at your table for chi-square and your results are not right: the square term and the (O-E)^2/E should all be positive terms. Try recalculating these terms to get a valid chi-square test statistic.

Also I'd recommend you also look at a regression analysis. If you are looking at testing say a model of Y = AX + B with H0: A = 0 vs H1: A != 0 then this can be done using standardized regression analysis with statistical software like R (which is free) or SAS (which is expensive).

You can also relate the chi-square results to the regression results to confirm that both answers are reasonable.

The regression can be used to add extra terms and you can test each co-efficient individually in a variety of different ways.

3. ## Re: Statistical analysis, this is impossible

The Chi squared test does not apply here. To apply the Chi squared test you need categorical data but the random variable you are testing is a continuous random variable. The groups you selected should be categories for the rate of CO2 generation, the temperature groups are just conditions not categories. Furthermore, the Chi squared statistic is calculated with expected and observed frequencies, not expected and observed values of CO2 generation.