I wish to study the trend of medical consultations each day during six years. Thus I expect near 2200 datas to analyse.
But some parameters are to consider :
- I don't have yet the datas per day, but the mean is about 2 consultations per day
- as it is difficult to do more than 3 or perhaps 4 consultations during one day (9h30 am - 13h pm), the others if they exist will probably be seen the next day (aso if the next day 3 news consultations occured)
- then, I don't know actually (as I expect the datas or each day but don't have now these datas) if the better distribution will be simply follow a Poisson' law, or exponential, or negative binomial, ..
- do someone have a clue for the better law given what i said ?
Further, I don't have a stastic program especially used for time trend, excepting Systran 13, but I don't believe that this program can be used with a theoric model of distribution, I recall that it does usual tasks, autocorrelations, saisonnal adjustments, .. but with continuous distributions I believe, and a linear model (removing the basic frequencies)
As such program (study of temporal series) is usually sold about 3000$ in France, that I don't expect to be a trader, with only one calculus to do, could anyone tell me how to adjust the better model to the 2200 datas that could be expected ?
Thx for your help, friendly yours;