I would be very grateful if sombody could help me how to analyze my data:
2 hospitals (one distant, one close) sends blood samples in my laboratory; I have the mean values of the analyte of interest for 48 months; I have the mean outside temperatures for that 48 months. I would like to know if the outside temperatures influence the mean values of my analyte (I presume that the samples arrive from the close hospital are not influenced, as there is little temperature change during transport, while the samples came from the distant hospital are influenced, as there are important temperature changes during transport). On basis of the graphical representation of my data (mean analyte concentrations and mean outside temperature through time (48 months) in the two hospitals) the analyte values are not influenced in non of the scenarios, but I need to demonstrate it also with a statistical test. But which one?
Many thanks!
This is a question for Jochen Wilhelm if he is willing to advice you.
Pearson's correlation coefficient test will serve the purpose as you have continuous variables of mean analyte concentrations and mean outside temperature, and you want to see the influence of temperature upon mean analyte concentrations, so it will show you how these two variables have association. You have large data for 48 months so it also fulfills the assumption of normal distribution.
I hope it will help
(Statstutor.ac.uk)
Than you both of you!
In the meantime I have made a linear regression analysis, for both hospitals, to see the correlation between temperature and mean analyte concentration. It seemed to me that regression analysis better fitted my scenario than the correlation test, but I am not really good in statistics.
Time series analysis should solve it. In contrast to regression this type of analysis adresses statistical dependence of time-sequenced data.
Simple linear regression analysis is also good as per your question because you have two quantitative variables and wanted to see the relationship between these two quantitative variables.
You can use the linear regression or logistic regression also, but if you use it, You have to make a variables some methods. (i.e, Dummy variables...etc. ) I think that there are many methods. In conclusion, you can use various methodologies depending on what you want to analyze.
A General Linear Model (GLM) ANOVA can start you off. This will confirm for you whether there are any significant differences between the two groups, then proceed to regression modeling if there are any.