Valeur en dessous du seuil de détection et analyse statistique sur sur JMP

[#1]

Bonjour,

Pour une étude, plusieurs valeurs (numériques continues) sont en dessous du seuil de détection.

Quelle est la méthode statistique recommandée pour exploiter ces données ?

Quelle application sur JMP ?

Merci beaucoup pour vos conseils.

CEB

[#2]

Hi @CEB87,

Welcome in the Community !

You can use Detection Limits in your response column property to define a lower and/or upper threshold beyond which a response cannot be measured.

When using Generalized Regression models (JMP Pro) in the Fit Model platform, this property (and corresponding values) will be used to create censored values when observation is oustside of these limits. You can watch the presentation Limits of Detection (LoD) - New in JMP Pro 16 to get a better understanding of the benefit to use detection limits in modeling.

Hope this answer will help you,

[#3]

Merci @Victor_G  

Cette fonctionnalité n'est pas disponible sur JMP 12. Une autre façon de faire? 

Merci 

CEB

[#4]

I'm not familiar with the options available in JMP12, but you could probably use Quantile Regression to deal with censored values in the Generalized Regression options.

The idea is either to use a regression model that doesn't estimate the conditional mean of the response variable across factors values, as it will be biased towards censored values (unlike estimating conditional median in Quantile Regression), and/or lower the importance of the censored data points on the model fit (to avoid the leverage effects of these censored values points).

Hope this answer will help you,