Data
Overview
For my analysis, I will be considering the rock type and rock amendment rate to be my predictor variables, as they are being controlled by me and varied across my sample pots. Further, while the rock type is a categorical variable, the amendment rate is a continuous variable (with units in kg/m2). My response variables, on the other hand, will be the nickel contamination rates for the collected leachates (in mg/L), as well as the alkalinity (in mM) of the leachates. Both response variables are continuous.
​
Figure 1. Header of datable used to compute exploratory statistics.
General Data Preparation
Nickel Contamination
In addition to the data preparation needed for each dataset which will be mentioned below, some general data preparation was needed in order to compute the exploratory graphics. This is due to the way that the various datasets were received/recorded. Each dataset originated in a different format/orientation, and the number of observations often varied between datasets due to sampling technique or errors in sampling/preparation. As such, many adjustments needed to be made when compiling the data to transpose or fill in the data as necessary such that one csv could be created for analysis in R. Further, unique IDs were created in excel to help with that data processing.
​
​
The nickel contamination data was prepared through a multistep process that accounted for changes in leachate volume across the experiment period, as well as mistakes in sample preparation.
​
Leachate volume correction: In total, across the experiment period, leachates were collected 5 times. However, the volume of leachate was not consistent for each collection. Thus, to get a representative concentration of nickel across the study period, a weighted average of the nickel concentrations needed to be taken, with weightings equivalent to the volume of leachate at each collection.
​
Improper sample preparation correction: In addition to the volume correction, the finalized nickel concentration was analyzed for errors that may have occurred due to improper sample filtration or acidification of the sample.
Figure 2. Nickel concentration in leachates corresponding to varying rock type/amendment rate combinations.
Alkalinity
The alkalinity data was prepared through a multistep process that accounted for changes in leachate volume across the experiment period, as well as mistakes in sample preparation.
​
Leachate volume correction: In total, across the experiment period, leachates were collected 5 times. However, the volume of leachate was not consistent for each collection. Thus, to get a representative alkalinity value across the study period, a weighted average of the alkalinities needed to be taken, with weightings equivalent to the volume of leachate at each collection.
​
Improper sample preparation correction: In addition to the volume correction, the finalized alkalinity values was analyzed for errors that may have occurred due to improper sample filtration or acidification of the sample.
Figure 3. Alkalinity of leachates corresponding to varying rock type/amendment rate combinations.
Initial Linear Models
Following the exploratory scatter plots, linear models were tested for individual rock types for both alkalinity and nickel concentration. This was done in are using the lm( ) function, and following the creation of the model residual plots were created to assess the fit. This was done for each of the five rock types, with the wollastonite examples provided below.
Figure 3. Linear model for the concentration of nickel in wollastonite leachates compared to amendment rate (left), with a corresponding residuals plot for that model (right).
Figure 3. Linear model for the alkalinity of wollastonite leachates compared to amendment rate (left), with a corresponding residuals plot for that model (right).