- Univariate Time Series
- Multiple Univariate Time Series
- Univariate Spatiotemporal Series
- Multiple Univariate Spatiotemporal Series
Univariate Time Series¶
In this case, we have a univariate time series.
For this, we will investigate how we can model a time series.
Tutorials:
- Global Mean Surface Temperature Anomaly
- Single Weather Station for Spain
Unconditional Density Estimation¶
D={yn}n=1N,N=NTyn∈RDy We have a few types of models we can use when we are faced with this situation.
Fully Pooled:Non-Pooled:Partially-Pooled:p(y,z,θ)p(y,z,θ)p(y,z,θ)=p(θ)p(z∣θ)n=1∏Np(yn∣z)=n=1∏Np(yn∣zn)p(zn∣θn)p(θn)=p(θ)n=1∏Np(yn∣zn)p(zn∣θ) The conclusion of this demonstration is that it’s almost always favourable to use a partially-pooled model as it effectively gives us options for modeling the dynamics of the parameters.
Temporally Conditioned Density Estimation¶
We can use conditional density estimation but we only condition on the time component.
In this case, we have some pairwise entries of measurements, yn, at some associated time stamp, tn.
D={tn,yn}n=1Nyn∈RDytn∈R+ where N=NT.
Measurements:Time Stamps:Latent Variables:Parameters:ytZθ∈RNT∈RNT∈RNT×Dz∈RDθyn∈Rtn∈R+zn∈RDz Here, we need to use a conditional
p(y,t,Z,θ)=p(θ)t=1∏NTp(yn∣zn)p(zn∣tn,θ)
Unconditional Dynamic Model¶
D={tn,yt}n=1Nyt∈RDyt∈R+ where N=NT.
Measurements:Latent Variables:Parameters:yZθ∈RNT∈RNT×Dz∈RDθyt∈Rzt∈RDz Finally, we can write the joint distribution
Strong-Constrained:Weak-Constrained:p(y,Z,θ)p(y,z,θ)=p(θ)p(z0∣θ)t=1∏Tp(yt∣zt)p(zt∣z0)=p(θ)p(z0∣θ)t=1∏Tp(yt∣zt)p(zt∣zt−1,θ)
Multivariate Time Series¶
In this case, we have a univariate time series.
For this, we will investigate how we can model a time series.
Tutorials:
- Single Weather Station for Spain + Multiple Variables
Unconditional Density Estimation¶
D={yn}n=1N,N=NTyn∈RDy We have a few types of models we can use when we are faced with this situation.
Partially-Pooled:p(Y,Z,θ)=p(θ)n=1∏Np(yn∣zn)p(zn∣θ) The conclusion of this demonstration is that it’s almost always favourable to use a partially-pooled model as it effectively gives us options for modeling the dynamics of the parameters.
Temporally Conditioned Density Estimation¶
We can use conditional density estimation but we only condition on the time component.
In this case, we have some pairwise entries of measurements, yn, at some associated time stamp, tn.
D={tn,yn}n=1Nyn∈RDytn∈R+ where N=NT.
Measurements:Time Stamps:Latent Variables:Parameters:YtZθ∈RNT×Dy∈RNT∈RNT×Dz∈RDθyn∈Rtn∈R+zn∈RDz Here, we need to use a conditional
p(Y,t,Z,θ)=p(θ)t=1∏NTp(yn∣zn)p(zn∣tn,θ)
Unconditional Dynamic Model¶
D={tn,yt}n=1Nyt∈RDyt∈R+ where N=NT.
Measurements:Latent Variables:Parameters:YZθ∈RNT×Dy∈RNT×Dz∈RDθyt∈RDyzt∈RDz Finally, we can write the joint distribution
Strong-Constrained:Weak-Constrained:p(Y,Z,θ)p(Y,Z,θ)=p(θ)p(z0∣θ)t=1∏Tp(yt∣zt)p(zt∣z0)=p(θ)p(z0∣θ)t=1∏Tp(yt∣zt)p(zt∣zt−1,θ)
Multivariate Spatiotemporal Series¶
In this case, we have a univariate time series.
For this, we will investigate how we can model a time series.
Tutorials:
- Multiple Weather Station for Spain + Multiple Variables
Unconditional Density Estimation¶
D={yn}n=1N,N=NTD=DyDΩyn∈RD We have a few types of models we can use when we are faced with this situation.
Partially-Pooled:p(Y,Z,θ)=p(θ)n=1∏Np(yn∣zn)p(zn∣θ) The conclusion of this demonstration is that it’s almost always favourable to use a partially-pooled model as it effectively gives us options for modeling the dynamics of the parameters.
Temporally Conditioned Density Estimation¶
Coordinate-Based¶
We can use conditional density estimation but we only condition on the time component.
In this case, we have some pairwise entries of measurements, yn, at some associated time stamp, tn.
D={(tn,sn),yn}n=1Nyn∈RDytn∈R+ where N=NT and D=Dy.
Measurements:Time Stamps:Spatial Coordinates:Latent Variables:Parameters:YtSZθ∈RNT×Dy∈RNT∈RNT×Ds∈RNT×Dz∈RDθyn∈RDytn∈R+sn∈RDszn∈RDz Here, we need to use a conditional
p(Y,t,S,Z,θ)=p(θ)t=1∏NTp(yn∣zn)p(zn∣tn,sn,θ)
Field-Based¶
We can use conditional density estimation but we only condition on the time component.
In this case, we have some pairwise entries of measurements, yn, at some associated time stamp, tn.
D={tn,yn}n=1Nyn∈RDtn∈R+ where N=NT and D=DΩDy.
Measurements:Time Stamps:Latent Variables:Parameters:YtZθ∈RNT×D∈RNT∈RNT×Dz∈RDθyn∈RDtn∈R+zn∈RDz Here, we need to use a conditional
p(Y,t,Z,θ)=p(θ)t=1∏NTp(yn∣zn)p(zn∣tn,θ)
Unconditional Dynamic Model¶
D={tn,yt}n=1Nyt∈RDt∈R+ where N=NT and D=DΩDy.
Measurements:Latent Variables:Parameters:YZθ∈RNT×D∈RNT×Dz∈RDθyt∈RDzt∈RDz Finally, we can write the joint distribution
Strong-Constrained:Weak-Constrained:p(Y,Z,θ)p(Y,Z,θ)=p(θ)p(z0∣θ)t=1∏Tp(yt∣zt)p(zt∣z0)=p(θ)p(z0∣θ)t=1∏Tp(yt∣zt)p(zt∣zt−1,θ)
Coupled Multivariate Spatiotemporal Series¶
In this case, we have a univariate time series.
For this, we will investigate how we can model a time series.
Tutorials:
- Multiple Weather Station for Spain + Multiple Variables
Conditional Density Estimation¶
D={xn,yn}n=1N,N=NTD=DyDΩyn∈RDxn∈RDx We have a few types of models we can use when we are faced with this situation.
Partially-Pooled:p(Y,X,Z,θ)=p(θ)n=1∏Np(yn∣zn)p(zn∣xn,θ) The conclusion of this demonstration is that it’s almost always favourable to use a partially-pooled model as it effectively gives us options for modeling the dynamics of the parameters.
Temporal Conditional Density Estimation¶
Coordinate-Based¶
We can use conditional density estimation but we only condition on the time component.
In this case, we have some pairwise entries of measurements, yn, at some associated time stamp, tn.
D={(tn,sn),xn,yn}n=1Nyn∈RDyxn∈RDxtn∈R+ where N=NT and D=Dy.
Measurements:Covariates:Time Stamps:Spatial Coordinates:Latent Variables:Parameters:YXtSZθ∈RNT×Dy∈RNT×Dx∈RNT∈RNT×Ds∈RNT×Dz∈RDθyn∈RDyxn∈RDxtn∈R+sn∈RDszn∈RDz Here, we need to use a conditional
p(Y,X,t,S,Z,θ)=p(θ)t=1∏NTp(yn∣zn)p(zn∣tn,sn,xn,θ)
Field-Based¶
We can use conditional density estimation but we only condition on the time component.
In this case, we have some pairwise entries of measurements, yn, at some associated time stamp, tn.
D={tn,xn,yn}n=1Nyn∈RDyxn∈RDxtn∈R+ where N=NT and Dy=DΩDy, Dx=DΩx.
Measurements:Covariates:Time Stamps:Latent Variables:Parameters:YXtZθ∈RNT×D∈RNT×Dx∈RNT∈RNT×Dz∈RDθyn∈RDxn∈RDxtn∈R+zn∈RDz Here, we need to use a conditional
p(Y,X,t,Z,θ)=p(θ)t=1∏NTp(yn∣zn)p(zn∣tn,xn,θ)
Conditional Dynamic Model¶
D={tn,xt,yt}n=1Nyt∈RDyxt∈RDxt∈R+ where N=NT and D=DΩDy.
Measurements:Covariates:Latent Variables:Parameters:YXZθ∈RNT×Dy∈RNT×Dx∈RNT×Dz∈RDθyt∈RDyxt∈RDxzt∈RDz Finally, we can write the joint distribution
Strong-Constrained:Weak-Constrained:p(Y,X,Z,θ)p(Y,X,Z,θ)=p(θ)p(z0∣θ)t=1∏Tp(yt∣zt)p(zt∣xt,z0)=p(θ)p(z0∣θ)t=1∏Tp(yt∣zt)p(zt∣xt,zt−1,θ)