# Missing value reconstruction based on ANOVA

### Description

Performs a simple missing value reconstruction based on an ANOVA with two factors using different methods.

### Usage

1 2 | ```
reconstruct(x, factor1.name, factor2.name,
data.name, method = "LES", iterations.num = 100)
``` |

### Arguments

`x` |
WUX data frame of class |

`factor1.name` |
Name of the 1st factor. |

`factor2.name` |
Name of the 2nd factor. |

`data.name` |
Name of the variable to be reconstructed. |

`method` |
One of the currently implemented methods: "LES", "Iterative" or "IterativeCC". See details section. |

`iterations.num` |
Number of iterations to be performed. Used only for |

### Details

Tools for filling missing values of an unbalanced climate model simulation matrix (e.g. missing RCM-GCM combinations of ENSEMBLES) in order to avoid biased ensemble estimates. Following methods are currently implemented:

`method = "LES"`

(default)

Performs a simple missing value reconstruction with two factors based on
solving the linear equation system (LES) of the ANOVA. The algorithm follows Déqué et al. (2007) but the
reconstruction is based on solving the linear equation system (LES) of
the ANOVA instead of reconstructing iteratively. The main advantages
of this method are that it is much faster and can be more easily
extended to more factors than the original one. However, keep in mind
that the results slightly differ from the iterative procedure proposed
by Déqué et al. (2007). The reconstruction algorithm is
based on unique factor combinations (i.e. one element per combination
of `factor1.name`

and `factor2.name`

).

`method = "Iterative"`

The data reconstruction follows the iterative procedure based on the
ANOVA proposed by Déqué et al. (2007). The reconstruction
algorithm is based on unique factor combinations (i.e. one element per
combination of `factor1.name`

and `factor2.name`

).

`method = "IterativeCC"`

Performs a leave one out cross calculation (CC) of the ANOVA based
missing value reconstruction with two factors based on and following the
iterative procedure of `method = "Iterative"`

.

### Value

Returns a WUX data frame containing the reconstructed data of class
`c("rwux.df", "wux.df", "data.frame")`

.

### Author(s)

Georg Heinrich g.heinrich@uni-graz.at and Thomas Mendlik thomas.mendlik@uni-graz.at

### References

Déqué M, Rowell DP, Lüthi D, Giorgi F, Christensen JH, Rockel B, Jacob D, Kjellström E, de Castro M, van den Hurk B. 2007. An intercomparison of regional climate simulations for Europe: Assessing uncertainties in model projections. Climatic Change 81: 53–70. DOI:10.1007/s10584-006-9228-x.

### Examples

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 | ```
## load WUX and read WUX test data
require(wux)
data(ensembles)
wuxtest.df <- subset(ensembles, subreg == "GAR")
## unique model acronyms are required for reconstruction
wuxtest.df$acronym <- factor(paste(wuxtest.df$institute, "_", wuxtest.df$rcm, sep=""))
## reconstruction of the missing data
reconstructLES.df <- reconstruct(wuxtest.df,
factor1.name = "acronym", factor2.name = "gcm", data.name =
"perc.delta.precipitation_amount", method = "LES")
## reconstruction of the missing data using iterative apporach from
## Deque et al (2007)
reconstructIterative.df <- reconstruct(wuxtest.df,
factor1.name = "acronym", factor2.name = "gcm", data.name =
"perc.delta.precipitation_amount", method = "Iterative",
iterations.num = 10)
## reconstruction of the missing data using iterative apporach with
## cross-calculation. This can take some time.
## Not run: reconstructIterative.df <- reconstruct(wuxtest.df,
factor1.name = "acronym", factor2.name = "gcm", data.name =
"perc.delta.precipitation_amount", method = "IterativeCC",
iterations.num = 10)
## End(Not run)
``` |