It is a macro which allows to select automatically fixed effects in PROC GLIMMIX.

You can choose the type of selection (BACKWARD or FORWARD), the threshold, until 5 level of random effects and force a variable in the model.

In order to algorithm converge, datafile must have not a lot of missing values

Adrien FRANCAIS

Macro to compare a quantitative variable to subgroups

Edited by Adrien FRANCAIS

adrien.francais@ujf-grenoble.fr

Macro to compare two groups according to a binary variable.

Effectives and Frequencies are computed for qualitative variables.

Distribution is described for quantitative variables.

Differences between groups is computed thanks to Khi2 test and Kruskall-Wallis test.

If the analyse is stratified, a term is available for that and pvalue appropriate is computed in logistic conditional regression.

Edited by Adrien FRANCAIS

adrien.francais@ujf-grenoble.fr

It is a macro which allows to decompose variables created in additional codes. We then obtain several variables with unique code to clearly identify value.

For example, if a variable is 19, we create 3 new variables with the code 16 for the first, 2 for the second and 1 for the last.

An example will help you.

Adrien FRANCAIS

It is a macro which allows to deconcatenate variable which contain multiple values separated by a character.

For example, if a variable is '674|675|676|677', we then create 4 new variables : 674 for the first, 675 for the second...until the last.

An example will help you.

Adrien FRANCAIS and Valérie SIROUX

Macro to describe cohort with frequencies, percentages and missing values for each modality of qualitative variables and also description of quantitative variables (mean, standard error, quartiles...) 

Edited by Adrien FRANCAIS

adrien.francais@ujf-grenoble.fr

Macro to compare distribution of quantitative and qualitative variables for a response variable which have several modalities.

There is no limit to the number of classes of response variable.

 An example will help you to apply this algorithm.

Edited by Adrien FRANCAIS

adrien.francais@ujf-grenoble.fr 

Designed by Muriel TAFFLET

Modified by Adrien FRANCAIS 

Macro which realizes a N:M matching according one or several qualitative variables

Designed by Aurélien VESIN

It is a macro which allows to transform qualitative variables in binary variables.

An example will help you.

Adrien FRANCAIS

It is a macro which allows to transform quantitative variables in classes.

You can choose the number of classes (2,4,5,10...)and the type of transformation : several binary variables according the percentile or only one new variable divided in 'n' classes.

An example will help you.

Adrien FRANCAIS and Aurélien VESIN

It is a macro which allows to rapidly validate a pronostic model thanks to discirmination (AUC and ROC curve), calibration (Hosmer-Lemeshow test and graph) and summarize quality of the model at the end.

You just must have probability of outcome and the outcome in a table.

This macro is very interesting for a validation dataset of a logistic model.

Adrien FRANCAIS