Impact of sampling and interpolation for time series
The raw data recorded when a change detected in the stream enables more realistic data recording comparison to time based sampling.Thus, sampling and interpolation required for most of the comparative analysis using time series, limit the usage of raw data for extended analysis. However, you might wonder how it can distort the realistic nature of the data (except the probable anomalies) and leading to false interpretation.Lets focus on our theme. For the analysis I am using fe w selected variables from water quality dataset recorded from Baffle Creek and Byrnett River . First I will combine the two datasets into single dataset without sampling using panads merge function. The Pearson Correlation Coefficient (PCC) calculated for the combined dataset and the heat-map of the results given below. Figure 1 : PCC between variables in two un-sampled datasets From the results, it is obvious that the two datasets may have correlations in-between the variable inside the dateset, thus, no