Description Usage Arguments Details Value Author(s) See Also Examples

This function allows estimating a mixture model of MCAR and MNAR values in each column of data sets similar to the ones which can be studied in MS-based quantitative proteomics. Such data matrices contain intensity values of identified peptides.

1 2 | ```
estim.mix(tab, tab.imp, conditions, x.step.mod=150, x.step.pi=150,
nb.rei=200)
``` |

`tab` |
A data matrix containing numeric and missing values. Each column of this matrix is assumed to correspond to an experimental sample, and each row to an identified peptide. |

`tab.imp` |
A matrix where the missing values of |

`conditions` |
A vector of factors indicating the biological condition to which each column (experimental sample) belongs. |

`x.step.mod` |
The number of points in the intervals used for estimating the cumulative distribution functions of the mixing model in each column. |

`x.step.pi` |
The number of points in the intervals used for estimating the proportion of MCAR values in each column. |

`nb.rei` |
The number of initializations of the minimization algorithm used to estimate the proportion of MCAR values (see Details). |

This function aims to estimate the following mixture model in each column:

*F_{tot}(x)=π_{na}\times F_{na}(x)+(1-π_{na})\times F_{obs}(x)*

*F_{na}(x)=π_{mcar}\times F_{tot}(x)+(1-π_{mcar})\times F_{mnar}(x)*

where *π_{na}* is the proportion of missing values, *π_{mcar}* is the proportion of MCAR values, *F_{tot}* is the cumulative distribution function (cdf) of the complete values, *F_{na}* is the cdf of the missing values, *F_{obs}* is the cdf of the observed values, and *F_{mnar}* is the cdf of the MNAR values.

To estimate this model, a first step consists to compute a rough estimate of *F_{na}* by assuming that all missing values are MCAR (thanks to the argument `tab.imp`

). This rough estimate is noted *\hat{F}_{na}*.

In a second step, the proportion of MCAR values is estimated. To do so, the ratio

*\hat{π}(x)=(1-\hat{F}_{na}(x))/(1-\hat{F}_{tot}(x))*

is computed for different *x*, where

*\hat{F}_{tot}(x)=π_{na}\times \hat{F}_{na}(x)+(1-π_{na})\times \hat{F}_{obs}(x)*

with *\hat{F}_{obs}* the empirical cdf of the observed values.

Next, the following minimization is performed:

*\min_{1>k>0,a>0,d>0}f(k,a,d)*

where

*f(k,a,d)=∑_x \frac{1}{s(x)^2}\times [\hat{π}(x)-k-(1-k)\frac{\exp(-a\times [x-lower_x]^d)}{1-\hat{F}_{tot}(x)}]^2*

where *s(x)^2* is an estimate of the asymptotic variance of *\hat{π}(x)*, *lower_x* is an estimate of the minimum of the complete values. To perform this minimization, the function `optim`

with the method "L-BFGS-B" is used. Because it is function of its initialization, it is possible to reinitialize a number of times the minimisation algorithm with the argument `nb.rei`

: the parameters leading to the lowest minimum are next kept.

Once `k`

, `a`

and `d`

are estimated, one can use several methods to estimate *π_{mcar}*: it is estimated
by *k*;

A list composed of:

`abs.pi` |
A numeric matrix containing the intervals used for estimating the ratio
in each column. |

`pi.init` |
A numeric matrix containing the estimated ratios
where |

`var.pi.init` |
A numeric matrix containing the estimated asymptotic variances of |

`trend.pi.init` |
A numeric matrix containing the estimated trend of the model used in the minimization algorithm. |

`abs.mod` |
A numeric vector containing the interval used for estimating the mixture models in each column. |

`pi.na` |
A numeric vector containing the proportions of missing values in each column. |

`F.na` |
A numeric matrix containing the estimated cumulative distribution functions of missing values in each column on the interval |

`F.tot` |
A numeric matrix containing the estimated cumulative distribution functions of complete values in each column on the interval |

`F.obs` |
A numeric matrix containing the estimated cumulative distribution functions of observed values in each column on the interval |

`pi.mcar` |
A numeric vector containing the estimations of the proportion of MCAR values in each column. |

`MinRes` |
A numeric matrix containing the three parameters of the model used in the minimization algorithm (three first rows), and the value of minimized function. |

Quentin Giai Gianetto <quentin2g@yahoo.fr>

1 2 3 4 5 6 7 8 | ```
#Simulating data
res.sim=sim.data(nb.pept=2000,nb.miss=600);
#Imputation of missing values with a MCAR-devoted algorithm: here the slsa algorithm
dat.slsa=impute.slsa(tab=res.sim$dat.obs,conditions=res.sim$condition,repbio=res.sim$repbio);
#Estimation of the mixture model
res=estim.mix(tab=res.sim$dat.obs, tab.imp=dat.slsa, conditions=res.sim$condition);
``` |

imp4p documentation built on Sept. 5, 2021, 5:38 p.m.

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.