I have historical and future climate data derived from different RCMs for different geographic locations. Each two groups of data are of unequal sizes. I did a Welch t-test on my samples and I sometimes find a significant difference between the two types of data and sometimes not!! Is that normal? and is it really necessary to test the significance of the data? I saw similar types of studies that did not do it.
If your question is regarding climate change (sorry, I guess it because I do not know what is acronym RCM), it is useful the estimation of trend and uncertainty.
https://chance.amstat.org/2017/11/climate-research/
Not necessary always. It's not needed in descriptive study or in qualitative methods. It depends on your objectives only.
Thank you Dr. Muhammed Ashraful Alam and Dr. Roberto Molteni : RCM is the acronym of Regional Climate Model. I am trying to evaluate the change in some climatic variables in the future.
Dr. Ette Etuk How to deal with the significance results that are changing from a sample to the other?
Quantity is not a matter in qualitative research. One single response might be better than 100 wrong answers and qualitative research can explain it in depth. Many things are beyond calculation and you are the best judge to decide according to your study context.
And of course qualitative or mixed method approach are based on science not on emotions only.
Dear Ghaieth Ben Hamouda
Thank you. When you use models, please keep in mind that is important a sensitivity and uncertainty analysis.
Dr Roberto Molteni I did it using a Welch t-test and I got different results some times for the same model but a different location
Dear Ghaieth Ben Hamouda ,
When I am referring to sensitivity analysis of the Regional Climate Model, I am referring to a study of how the uncertainty in the output of the model can be apportioned to different sources of uncertainty in the model input.
The uncertainty in the output of the model should be showed and discussed.
Dr Roberto Molteni What methods do you suggest to study the uncertainty?
Dear Ghaieth Ben Hamouda
Every field of science has its own specificity; however, Monte Carlo simulation is one of the common tools.
Take into account the possibility to separating Variability and Uncertainty, because they have different impact on the model output and model improvement.
Variability is part of the “natural” variation, e.g. variation of body weight; Uncertainty is our lack of knowledge. Variation can only be monitored, Uncertainty can be reduced.
https://ec.europa.eu/jrc/en/samo
and
Andrea Saltelli
http://www.andreasaltelli.eu/
He has worked at the European Commission, leading a unit of econometrics and applied statistics.
Using the significance test is not always necessary for all researches. Qualitative researches, for example, does not need a significance test. As for the quantitative researches, it depends on the nature of the research.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis