A full bivariate investigation

The scatterplot shows a strong positive linear association

The scatterplot shows a weak positive linear association

The scatterplot shows a moderate positive linear association

The scatterplot shows a moderate negative linear association

The residuals are irregularly scattered on both sides of zero, so the residual evidence supports the use of a linear model

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals include both positive and negative values, so they cancel and show that the fitted line predicts every response exactly

floor area is used to predict quarterly electricity use; 49.0% of variation in quarterly electricity use is explained through the fitted linear model, the irregular residual scatter supports the linear model, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

floor area is used to predict quarterly electricity use; 49.0% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

quarterly electricity use is used to predict floor area; 49.0% of variation in floor area is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. This observational analysis does not establish causation

floor area is used to predict quarterly electricity use; 49.0% of variation in quarterly electricity use is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. This observational analysis does not establish causation

floor area is used to predict quarterly electricity use; 49.0% of variation in quarterly electricity use is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. Therefore changing floor area causes quarterly electricity use to change

waste collected is the explanatory variable and event attendance is the response variable

event attendance is the explanatory variable and waste collected is the response variable

The scatterplot shows a moderate positive linear association

The scatterplot shows a moderate negative linear association

The scatterplot shows a weak positive linear association

The scatterplot shows a strong positive linear association

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals include both positive and negative values, so they cancel and show that the fitted line predicts every response exactly

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

The residuals are irregularly scattered on both sides of zero, so the residual evidence supports the use of a linear model

waste collected is used to predict event attendance; 43.8% of variation in event attendance is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. This observational analysis does not establish causation

event attendance is used to predict waste collected; 43.8% of variation in waste collected is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. This observational analysis does not establish causation

event attendance is used to predict waste collected; 43.8% of variation in waste collected is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. Therefore changing event attendance causes waste collected to change

event attendance is used to predict waste collected; 43.8% of variation in waste collected is explained through the fitted linear model, the irregular residual scatter supports the linear model, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

event attendance is used to predict waste collected; 43.8% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

processing time is the explanatory variable and parcels handled is the response variable

parcels handled is the explanatory variable and processing time is the response variable

The scatterplot shows a strong positive linear association

The scatterplot shows a moderate negative linear association

The scatterplot shows a weak positive linear association

The scatterplot shows a moderate positive linear association

Across the displayed bivariate data, predicted processing time increases by about 2.0 on the displayed min scale in total; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in parcels handled (parcels), predicted processing time increases by about 2.0 on the displayed min scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in parcels handled (parcels), the slope 2.0 proves the association between the two plotted variables is strong; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in parcels handled (parcels), predicted processing time increases by about 2.0 on the displayed min scale; the intercept at EV = 0 is a reliable prediction

For each 1-unit increase in processing time (min), predicted parcels handled increases by about 0.2 on the displayed parcels scale; the intercept at EV = 0 is outside the observed range

The residuals are irregularly scattered on both sides of zero, so the residual evidence supports the use of a linear model

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

The residuals include both positive and negative values, so they cancel and show that the fitted line predicts every response exactly

parcels handled is used to predict processing time; 42.2% of variation in processing time is explained through the fitted linear model, the irregular residual scatter supports the linear model, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

parcels handled is used to predict processing time; 42.2% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

parcels handled is used to predict processing time; 42.2% of variation in processing time is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. This observational analysis does not establish causation

processing time is used to predict parcels handled; 42.2% of variation in parcels handled is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. This observational analysis does not establish causation

parcels handled is used to predict processing time; 42.2% of variation in processing time is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. Therefore changing parcels handled causes processing time to change

practice time is the explanatory variable and practice points is the response variable

practice points is the explanatory variable and practice time is the response variable

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals are irregularly scattered on both sides of zero, so the residual evidence supports the use of a linear model

The residuals include both positive and negative values, so they cancel and show that the fitted line predicts every response exactly

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

practice time is used to predict practice points; 42.3% of variation in practice points is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. This observational analysis does not establish causation

practice time is used to predict practice points; 42.3% of variation in practice points is explained through the fitted linear model, the irregular residual scatter supports the linear model, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

practice time is used to predict practice points; 42.3% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

practice time is used to predict practice points; 42.3% of variation in practice points is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. Therefore changing practice time causes practice points to change

practice points is used to predict practice time; 42.3% of variation in practice time is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an interpolation supported by the observed explanatory range. This observational analysis does not establish causation

For each 1-unit increase in machine age (months), predicted hourly output decreases by about 1.2 on the displayed units scale; the intercept at EV = 0 is a reliable prediction

For each 1-unit increase in machine age (months), the slope 1.2 proves the association between the two plotted variables is strong; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in hourly output (units), predicted machine age decreases by about 0.4 on the displayed months scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in machine age (months), predicted hourly output decreases by about 1.2 on the displayed units scale; the intercept at EV = 0 is outside the observed range

Across the displayed bivariate data, predicted hourly output decreases by about 1.2 on the displayed units scale in total; the intercept at EV = 0 is outside the observed range

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

The residuals include both positive and negative values, so they cancel and show that the fitted line predicts every response exactly

The residuals are irregularly scattered on both sides of zero, so the residual evidence supports the use of a linear model

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

hourly output is used to predict machine age; 48.1% of variation in machine age is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

machine age is used to predict hourly output; 48.1% of variation in hourly output is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

machine age is used to predict hourly output; 48.1% of variation in hourly output is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an extrapolation not supported by the observed explanatory range. Therefore changing machine age causes hourly output to change

machine age is used to predict hourly output; 48.1% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

machine age is used to predict hourly output; 48.1% of variation in hourly output is explained through the fitted linear model, the irregular residual scatter supports the linear model, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

For each 1-unit increase in machine age (months), the slope 4.00 proves the association between the two plotted variables is strong; the intercept at EV = 0 is outside the observed range

Across the displayed bivariate data, predicted hourly output decreases by about 4.00 on the displayed units scale in total; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in machine age (months), predicted hourly output decreases by about 4.00 on the displayed units scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in hourly output (units), predicted machine age decreases by about 0.22 on the displayed months scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in machine age (months), predicted hourly output decreases by about 4.00 on the displayed units scale; the intercept at EV = 0 is a reliable prediction

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals include both positive and negative values, so their systematic curved pattern still supports the use of a linear model

The residuals show a systematic curved pattern around zero, so the residual evidence calls the linear model into question

machine age is used to predict hourly output; 86.5% of variation in hourly output is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

machine age is used to predict hourly output; 86.5% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

hourly output is used to predict machine age; 86.5% of variation in machine age is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

machine age is used to predict hourly output; 86.5% of variation in hourly output is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

machine age is used to predict hourly output; 86.5% of variation in hourly output is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. Therefore changing machine age causes hourly output to change

For each 1-unit increase in elevation (hundreds of metres), predicted air pressure decreases by about 4.0 on the displayed hPa scale; the intercept at EV = 0 is a reliable prediction

Across the displayed bivariate data, predicted air pressure decreases by about 4.0 on the displayed hPa scale in total; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in air pressure (hPa), predicted elevation decreases by about 0.2 on the displayed hundreds of metres scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in elevation (hundreds of metres), predicted air pressure decreases by about 4.0 on the displayed hPa scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in elevation (hundreds of metres), the slope 4.0 proves the association between the two plotted variables is strong; the intercept at EV = 0 is outside the observed range

The residuals show a systematic curved pattern around zero, so the residual evidence calls the linear model into question

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals include both positive and negative values, so their systematic curved pattern still supports the use of a linear model

elevation is used to predict air pressure; 88.9% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

elevation is used to predict air pressure; 88.9% of variation in air pressure is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

elevation is used to predict air pressure; 88.9% of variation in air pressure is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

air pressure is used to predict elevation; 88.9% of variation in elevation is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

elevation is used to predict air pressure; 88.9% of variation in air pressure is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. Therefore changing elevation causes air pressure to change

heating energy use is the explanatory variable and outdoor temperature is the response variable

outdoor temperature is the explanatory variable and heating energy use is the response variable

For each 1-unit increase in outdoor temperature (°C), the slope 1.14 proves the association between the two plotted variables is strong; the intercept at EV = 0 is outside the observed range

Across the displayed bivariate data, predicted heating energy use decreases by about 1.14 on the displayed MJ scale in total; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in heating energy use (MJ), predicted outdoor temperature decreases by about 0.41 on the displayed °C scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in outdoor temperature (°C), predicted heating energy use decreases by about 1.14 on the displayed MJ scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in outdoor temperature (°C), predicted heating energy use decreases by about 1.14 on the displayed MJ scale; the intercept at EV = 0 is a reliable prediction

The residuals include both positive and negative values, so they cancel and show that the fitted line predicts every response exactly

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

The residuals are irregularly scattered on both sides of zero, so the residual evidence supports the use of a linear model

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

outdoor temperature is used to predict heating energy use; 47.0% of variation in heating energy use is explained through the fitted linear model, the irregular residual scatter supports the linear model, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

outdoor temperature is used to predict heating energy use; 47.0% of variation in heating energy use is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

outdoor temperature is used to predict heating energy use; 47.0% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

heating energy use is used to predict outdoor temperature; 47.0% of variation in outdoor temperature is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

outdoor temperature is used to predict heating energy use; 47.0% of variation in heating energy use is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an extrapolation not supported by the observed explanatory range. Therefore changing outdoor temperature causes heating energy use to change

trip distance is the explanatory variable and remaining battery range is the response variable

remaining battery range is the explanatory variable and trip distance is the response variable

For each 1-unit increase in trip distance (km), the slope 4.0 proves the association between the two plotted variables is strong; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in trip distance (km), predicted remaining battery range decreases by about 4.0 on the displayed km scale; the intercept at EV = 0 is outside the observed range

Across the displayed bivariate data, predicted remaining battery range decreases by about 4.0 on the displayed km scale in total; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in remaining battery range (km), predicted trip distance decreases by about 0.2 on the displayed km scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in trip distance (km), predicted remaining battery range decreases by about 4.0 on the displayed km scale; the intercept at EV = 0 is a reliable prediction

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

The residuals show a systematic curved pattern around zero, so the residual evidence calls the linear model into question

The residuals include both positive and negative values, so their systematic curved pattern still supports the use of a linear model

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

remaining battery range is used to predict trip distance; 85.1% of variation in trip distance is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

trip distance is used to predict remaining battery range; 85.1% of variation in remaining battery range is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. Therefore changing trip distance causes remaining battery range to change

trip distance is used to predict remaining battery range; 85.1% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

trip distance is used to predict remaining battery range; 85.1% of variation in remaining battery range is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

trip distance is used to predict remaining battery range; 85.1% of variation in remaining battery range is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

machine age is the explanatory variable and hourly output is the response variable

hourly output is the explanatory variable and machine age is the response variable

For each 1-unit increase in machine age (months), predicted hourly output decreases by about 1.0 on the displayed units scale; the intercept at EV = 0 is a reliable prediction

For each 1-unit increase in hourly output (units), predicted machine age decreases by about 0.8 on the displayed months scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in machine age (months), predicted hourly output decreases by about 1.0 on the displayed units scale; the intercept at EV = 0 is outside the observed range

For each 1-unit increase in machine age (months), the slope 1.0 proves the association between the two plotted variables is strong; the intercept at EV = 0 is outside the observed range

Across the displayed bivariate data, predicted hourly output decreases by about 1.0 on the displayed units scale in total; the intercept at EV = 0 is outside the observed range

The residuals show a systematic curved pattern around zero, so the residual evidence calls the linear model into question

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals include both positive and negative values, so their systematic curved pattern still supports the use of a linear model

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

machine age is used to predict hourly output; 77.8% of variation in hourly output is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

machine age is used to predict hourly output; 77.8% of variation in hourly output is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. Therefore changing machine age causes hourly output to change

hourly output is used to predict machine age; 77.8% of variation in machine age is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

machine age is used to predict hourly output; 77.8% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

machine age is used to predict hourly output; 77.8% of variation in hourly output is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

soil moisture reading is the explanatory variable and time since watering is the response variable

time since watering is the explanatory variable and soil moisture reading is the response variable

The residuals are irregularly scattered on both sides of zero, so the residual evidence supports the use of a linear model

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals include both positive and negative values, so they cancel and show that the fitted line predicts every response exactly

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

time since watering is used to predict soil moisture reading; 42.0% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

soil moisture reading is used to predict time since watering; 42.0% of variation in time since watering is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

time since watering is used to predict soil moisture reading; 42.0% of variation in soil moisture reading is explained through the fitted linear model, the irregular residual scatter supports the linear model, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

time since watering is used to predict soil moisture reading; 42.0% of variation in soil moisture reading is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

time since watering is used to predict soil moisture reading; 42.0% of variation in soil moisture reading is explained through the fitted linear model, the irregular residual scatter supports the linear model, and the queried value is an extrapolation not supported by the observed explanatory range. Therefore changing time since watering causes soil moisture reading to change

trip distance is the explanatory variable and remaining battery range is the response variable

remaining battery range is the explanatory variable and trip distance is the response variable

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

The residuals show a systematic curved pattern around zero, so the residual evidence calls the linear model into question

The residuals include both positive and negative values, so their systematic curved pattern still supports the use of a linear model

trip distance is used to predict remaining battery range; 86.9% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

remaining battery range is used to predict trip distance; 86.9% of variation in trip distance is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

trip distance is used to predict remaining battery range; 86.9% of variation in remaining battery range is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

trip distance is used to predict remaining battery range; 86.9% of variation in remaining battery range is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. Therefore changing trip distance causes remaining battery range to change

trip distance is used to predict remaining battery range; 86.9% of variation in remaining battery range is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

soil moisture reading is the explanatory variable and time since watering is the response variable

time since watering is the explanatory variable and soil moisture reading is the response variable

The response values should be on the horizontal axis instead of the explanatory values, so this residual plot cannot assess the linear model

The residuals include both positive and negative values, so their systematic curved pattern still supports the use of a linear model

The residuals show a systematic curved pattern around zero, so the residual evidence calls the linear model into question

The residuals do not all lie exactly on the zero line, so the residual evidence always rules out the use of a linear model

time since watering is used to predict soil moisture reading; 80.5% of variation in soil moisture reading is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

time since watering is used to predict soil moisture reading; 80.5% explained variation is treated as enough by itself to prove the fitted linear model is appropriate even when residual evidence conflicts, and the queried value is treated as supported at every explanatory value. This observational analysis does not establish causation

soil moisture reading is used to predict time since watering; 80.5% of variation in time since watering is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. This observational analysis does not establish causation

time since watering is used to predict soil moisture reading; 80.5% of variation in soil moisture reading is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, but the model is then used at every explanatory value without checking the observed range. This observational analysis does not establish causation

time since watering is used to predict soil moisture reading; 80.5% of variation in soil moisture reading is explained through the fitted linear model, the systematic residual pattern calls the linear model into question, and the queried value is an extrapolation not supported by the observed explanatory range. Therefore changing time since watering causes soil moisture reading to change