An automatic report for the dataset : Pheonix Housing Price
Abstract
This report was produced by the Automatic Bayesian Covariance Discovery (ABCD) algorithm.
1 Executive summary
The raw data and full model posterior with extrapolations are shown in figure 1.
The structure search algorithm has identified five additive components in the data. The first 2 additive components explain 99.2% of the variation in the data as shown by the coefficient of determination (${R}^{2}$) values in table 1. After the first 4 components the cross validated mean absolute error (MAE) does not decrease by more than 0.1%. This suggests that subsequent terms are modelling very short term trends, uncorrelated noise or are artefacts of the model or search procedure. Short summaries of the additive components are as follows:

•
A constant.

•
A smooth function.

•
A smooth function with linearly decreasing marginal standard deviation.

•
A smooth function.

•
Uncorrelated noise.
#  ${R}^{2}$ (%)  $\mathrm{\Delta}{R}^{2}$ (%)  Residual ${R}^{2}$ (%)  Cross validated MAE  Reduction in MAE (%) 

        148.29   
1  0.0  0.0  0.0  36.51  75.4 
2  99.2  99.2  99.2  3.91  89.3 
3  100.0  0.8  99.2  2.08  46.7 
4  100.0  0.0  88.0  2.06  1.0 
5  100.0  0.0  100.0  2.06  0.0 
Model checking statistics are summarised in table 2 in section 4. These statistics have not revealed any inconsistencies between the model and observed data.
The rest of the document is structured as follows. In section 2 the forms of the additive components are described and their posterior distributions are displayed. In section 3 the modelling assumptions of each component are discussed with reference to how this affects the extrapolations made by the model. Section 4 discusses model checking statistics, with plots showing the form of any detected discrepancies between the model and observed data.
2 Detailed discussion of additive components
2.1 Component 1 : A constant
This component is constant.
This component explains 0.0% of the total variance. The addition of this component reduces the cross validated MAE by 75.4% from 148.3 to 36.5.
2.2 Component 2 : A smooth function
This component is a smooth function with a typical lengthscale of 1.5 years.
This component explains 99.2% of the residual variance; this increases the total variance explained from 0.0% to 99.2%. The addition of this component reduces the cross validated MAE by 89.30% from 36.51 to 3.91.
2.3 Component 3 : A smooth function with linearly decreasing marginal standard deviation
This component is a smooth function with a typical lengthscale of 4.6 months. The marginal standard deviation of the function decreases linearly.
This component explains 99.2% of the residual variance; this increases the total variance explained from 99.2% to 100.0%. The addition of this component reduces the cross validated MAE by 46.69% from 3.91 to 2.08.
2.4 Component 4 : A smooth function
This component is a smooth function with a typical lengthscale of 6.6 weeks.
This component explains 88.0% of the residual variance; this increases the total variance explained from 100.0% to 100.0%. The addition of this component reduces the cross validated MAE by 0.99% from 2.08 to 2.06.
2.5 Component 5 : Uncorrelated noise
This component models uncorrelated noise.
This component explains 100.0% of the residual variance; this increases the total variance explained from 100.0% to 100.0%. The addition of this component reduces the cross validated MAE by 0.00% from 2.06 to 2.06. This component explains residual variance but does not improve MAE which suggests that this component describes very short term patterns, uncorrelated noise or is an artefact of the model or search procedure.
3 Extrapolation
Summaries of the posterior distribution of the full model are shown in figure 11. The plot on the left displays the mean of the posterior together with pointwise variance. The plot on the right displays three random samples from the posterior.
Below are descriptions of the modelling assumptions associated with each additive component and how they affect the predictive posterior. Plots of the pointwise posterior and samples from the posterior are also presented, showing extrapolations from each component and the cuulative sum of components.
3.1 Component 1 : A constant
This component is assumed to stay constant.
3.2 Component 2 : A smooth function
This component is assumed to continue smoothly but is also assumed to be stationary so its distribution will return to the prior. The prior distribution places mass on smooth functions with a marginal mean of zero and a typical lengthscale of 1.5 years. [This is a placeholder for a description of how quickly the posterior will start to resemble the prior].
3.3 Component 3 : A smooth function with linearly decreasing marginal standard deviation
This component is assumed to continue smoothly but is also assumed to be stationary so its distribution will return to the prior. The prior distribution places mass on smooth functions with a marginal mean of zero and a typical lengthscale of 4.6 months. [This is a placeholder for a description of how quickly the posterior will start to resemble the prior]. The marginal standard deviation of the function is assumed to continue to decrease linearly until May 2015 after which the marginal standard deviation of the function is assumed to start increasing linearly.
3.4 Component 4 : A smooth function
This component is assumed to continue smoothly but is also assumed to be stationary so its distribution will return to the prior. The prior distribution places mass on smooth functions with a marginal mean of zero and a typical lengthscale of 6.6 weeks. [This is a placeholder for a description of how quickly the posterior will start to resemble the prior].
3.5 Component 5 : Uncorrelated noise
This component assumes the uncorrelated noise will continue indefinitely.
4 Model checking
Several posterior predictive checks have been performed to assess how well the model describes the observed data. These tests take the form of comparing statistics evaluated on samples from the prior and posterior distributions for each additive component. The statistics are derived from autocorrelation function (ACF) estimates, periodograms and quantilequantile (qq) plots.
Table 2 displays cumulative probability and $p$value estimates for these quantities. Cumulative probabilities near 0/1 indicate that the test statistic was lower/higher under the posterior compared to the prior unexpectedly often i.e. they contain the same information as a $p$value for a twotailed test and they also express if the test statistic was higher or lower than expected. $p$values near 0 indicate that the test statistic was larger in magnitude under the posterior compared to the prior unexpectedly often.
ACF  Periodogram  
#  min  min loc  max  max loc  max  min 
1  0.487  0.517  0.616  0.506  0.166  0.834 
2  0.639  0.518  0.873  0.550  0.494  0.797 
3  0.254  0.308  0.795  0.794  0.602  0.544 
4  0.438  0.494  0.510  0.503  0.490  0.515 
5  0.427  0.493  0.277  0.240  0.406  0.464 
No statistically significant discrepancies between the data and model have been detected but model checking plots for each component are presented below.
4.1 Model checking plots for components without statistically significant discrepancies
4.1.1 Component 1 : A constant
No discrepancies between the prior and posterior of this component have been detected
4.1.2 Component 2 : A smooth function
No discrepancies between the prior and posterior of this component have been detected
4.1.3 Component 3 : A smooth function with linearly decreasing marginal standard deviation
No discrepancies between the prior and posterior of this component have been detected
4.1.4 Component 4 : A smooth function
No discrepancies between the prior and posterior of this component have been detected
4.1.5 Component 5 : Uncorrelated noise
No discrepancies between the prior and posterior of this component have been detected
5 MMD  experimental section
#  mmd 

1  0.000 
2  0.000 
3  0.000 
4  0.019 
5  0.606 