Why Use This Data Source In Your Models?
Weather is utilized in order to provide a measure of behavioral changes based on variations. This can include both severe weather as well as overall shifts in weather as a dynamic form of seasonality.
The data shows auto correlation and a non-normal distribution. The data should be differenced. While the Square Root transformation, provides the best normality, the Arcsin variable will also perform well.
Data is able to be distributed by time but not by geography. The roll up method used is Weighted Average.
Data shows auto correlation indicating a need for differencing
The ACF indicates 1 order differencing is appropriate.
Following first order differencing, no further differencing is required based on the differenced ACF at lag one of -0.39
The Kwiatkowski-Phillips-Schmidt-Shin (KPSS) test, KPSS Trend = 0.14 p-value = 0.07 indicates that the data is stationary.
The Shapiro-Wilk test returned W = 0.30 with a p-value =0.00 indicating the data does not follow a normal distribution.
A skewness score of 5.92 indicates the data are substantially skewed.
Hartigan's dip test score of 0.02 with a p-value of 0.00 inidcates the data is multimodal
Auto Correlation Function
Auto Correlation Function After Differencing
Partial Auto Correlation Function
Seasonal and Trend Decompostion
Some weather stations, such as the State of Delaware, do not report as frequently as others.
Menne, M.J., I. Durre, B. Korzeniewski, S. McNeal, K. Thomas, X. Yin, S. Anthony, R. Ray, R.S. Vose, B.E.Gleason, and T.G. Houston, 2012: Global Historical Climatology Network - Daily (GHCN-Daily), Version 3.27; NOAA National Climatic Data Center. http://doi.org/10.7289/V5D21VHZ [access date].
Use our platform to aggregate, normalize, and profile open source and premium control data. Spend less time finding and wrangling data, and more time building efficient and feature-rich machine learning data pipelines.
Instantly apply industry-standard
data science treatments and transformations, including (but not limited to) Differencing, Lead/Lag, Box Cox. Easily manipulate data across different time and geographic grains.
Our Patent Pending iterative testing engine allows you to upload your target variable, and the platform will test for possible statistical relationships across all available data sources. Saving you time and removing analyst bias.
Easily integrate your Ready Signal data to the data science platform of your choice. Connect directly to Ready Signal through our API or using one of our pre-built data connectors or download directly in Excel or CSV format.