Dear Julie

My recommend try to search the methods that introduced by Jacobsen et al., and Wasserburg et al., at year 1980's where both scientist are from Harvard.  Possibly  it might  help your research

Thank you,

Best regard's

Zolhizir

Dear Julie,

If this kind of data does not have any specific particularities I believe an acceptable method would be to simply average the data at higher resolution to be equivalent to the lower resolution one. I'm assuming that your data at lower resolution actually represent an average of the period covered by some data points on the higher resolution data. As both your data have chronological determination, this should be quite straightforward. If the points in the higher resolution data do not represent equal amounts of time, use weighted averages. But if in your data the point do not cover the period of the data points from the higher resolution, there are a few alternatives. You could simply select only the data points in the higher resolution data that cover equivalent periods of your data if that is possible. If not, you will probably need a statistical model and interpolation.

If the objective is trend comparison, perhaps you do not need to solve any differences on data resolution. It is easy to compare linear trends for example, testing for differences on the angular coefficient of two regression models, but don't forget that hypothesis testing is a trick thing and was developed in the context of experimental data, where delineation was defined by the researcher and "more extreme than your data" (tail probability) has a clear meaning. But I understand that statistical tests are required by most journals, so to help it to be meaningful you have at least to perform a power analysis, where you have to define (personal judgment) the minimum difference that would be geophysically significant (in contrast to statistically significant). Also, always report your estimates and add some measure of uncertainty (e.g., confidence bounds) independently of the result of the test (you estimates are real, test result are mathematical constructs).

Dear Zolhizir bin Daud and Alexandre Castagna,

thank you both for your help.It has become  already clearer for me.

Indeed, my data at lower resolution covers exactly the period of the data at higher resolution. However, the data at higher resolution is not evenly distributed. At the moment, the software I am using (Sigma plot) do not have a weighted average, but I can change the option to obtain a uniformely spaced values using the nearest neighbour bandwidth method. I can also match the interval number with the number of data point of my own low resolution data. Doing that, I have visualy compared my terrestrial data with the benthic data and find that they don't match very well, meaning that my site is preobably effected by other factors than global temperatures (to be determined). However, as Alexandre Castagna pointed, it is good to show statistics. I understand the principle that you describe, now I have to work out how to do that with my software. I guess now is the time to switch to R or SPSS...

Thank you for your guidance,

Julie

Very nice Julie,

I would suggest to you R as very flexible and powerful tool and you can always count on its large community of users to help you out. There are a number of e-mail groups dedicated to specific research areas with R. Good luck with your research.