I have data of rainfall for 6 years.It consist of mostly zero values.I checked the normality of the data and it show abnormal distribution.I want to know how can i make it normal without reducing it's size.
You can consult this:
How to analyse rainfall data - Hydrology Project
https://www.mahahp.gov.in › resource › modules
by DHVC BV · Cited by 3 — Use normal distribution fitting for maximum monthly rainfall data series.
Hi
You can use non-parametric tests. In this case, there is no need to change the size of the data. I do not know another way. I hope it helped.
Hi Iram,
Precipitation is a skewed distribution so you will need to transform it to something that approaches more to a normal one. You may want to use a square root transformation. This way you avoid some problems with zeros you have when you use log transformations.
After the transformation the data will look more like normal, but do not expect a perfectly normal distribution. Then you work on the transformed space and when you finish you transform back to the original space. Depending on what you do you should be careful on the back transformation. We have applied square root transformation on the CombiPrecip project in Switzerland and it has worked well until now for our purposes (Sideris, I.V., Gabella, M., Erdin, R. and Germann, U., 2014. Real‐time radar–rain‐gauge merging using spatio‐temporal co‐kriging with external drift in the alpine terrain of Switzerland. Quarterly Journal of the Royal Meteorological Society, 140(680), pp.1097-1111.).
On the other hand, if you do not care so much about the computational cost/time of the transformation and you want to end up to a perfectly normal distribution you may want to check section 7.2.2 "Normal score trasnform" in the book by Pierre Goovaerts "Geostratistics for natural resources evaluation". In Figure 7.3 in this book you can immediately see the result.
Good luck with your project.
May want to use a two step approach as done with the CLIGEN weather database. The first step is the probability of of a wet day following a wet day and a wet day following a dry day for each month. The second step is the mean daily precipitation on a wet day. The distribution of the daily precipitation will not be normal, so a transform will be necessary. Use a normality analysis for the linear data and for some typical transforms, like log, square root etc, and find out which transform results in the best normality value. SAS, R, and other stats packages have normality tests which are described on Google. Very few hydrologic data sets are normally distributed.
You need to difference all the values in your time series. It is better if values are divided by the mean value of the time series.
You can transform the data. As there are many zero values, be careful when you transform.
you should find outliers, maybe you don't need to tranformation the data.
- Badges
- Science topic
- Similar topics
- Filing