Les graphiques : un sujet pour qui ?

En recherche clinique, les biostatisticiens sont les principaux utilisateurs de graphiques. Ils servent à illustrer leurs propos auprès de multiples acteurs ayant une expertise technique très variable. Ces derniers peuvent également être les fournisseurs ou déléguer cette tâche aux programmeurs. Les langages de programmation come SAS et R restent les langages privilégiés pour ces analyses.

L’atout majeur, selon moi, est de pouvoir combiner plusieurs bénéfices : la traçabilité de l’information, la reproductibilité des résultats, la réutilisabilité des précédentes productions, une certaine flexibilité pour s’adapter aux particularités d’une étude et la possibilité d’utiliser des statistiques plus avancées.

Les autres fonctions peuvent également bénéficier de la visualisation. Ainsi les managers peuvent assurer un suivi des études cliniques par un dashboard avec visualisation, qu’il s’agisse du planning des études ou de l’analyse de paramètres (nombre de queries…). Ils peuvent aussi avoir des reporting sur les performances des programmes, des systèmes informatiques, etc. Les DM programmeurs peuvent compléter la vérification des données par des graphiques illustrant des patterns destinés aux medical experts, les incohérences étant difficilement détectables autrement.

Les alternatives comme PowerBI, Tableau ou Spotfire sont designées pour rapporter une sélection de statistiques sous forme de dashboard. Elles vendent un état des lieux pouvant inclure des notions temporelles (statistique au temps t, sur une période donnée) et des comparaisons. Elles répondent à une demande d’un public ayant un besoin statistique moins technique, à une recherche de statistiques facilement compréhensibles et interpréables. La mise en oeuvre, moins technique, rend la réalisation des graphiques plus abordable. Celui qui produit les graphiques et celui qui les utilisent peuvent être une seule et même personne. Ces alternatives favorisent l’autonomie des utilisateurs, automatie souvent recherchée, et réduisent les zones de flou dans la communication entre professionnels venant d’horizons différents.

Excel, utilisé pour un usage ponctuel, n’est pas le plus indiqué pour l’automatisation, le contrôle qualité et la traçabilité des données. Il peut être un point de départ d’une réflexion, voire d’une analyse exploratoire. Il est souvent le seul outil disponible aux non-techniciens, tels que les experts médicaux, pour observer les données grâce à la visualisation.

Comme nous venons de le voir chacune de ces approches a ses avantages et ses inconvénients.

Je fais le choix, ici, de vous parler uniquement des graphiques produits en SAS, langage historique toujours utilisé par de nombreux acteurs de la recherche clinique.

Customized graphics with proc sgplot made fun and intuitive

• 7x 90-minute sessions: follow the 7 steps to customizing graphics
• 20 guided exercises in your inbox: reinforce your skills through practice
• GoGoGraph monthly community meeting: share your experiences and discuss your current graph challenges

Why starting with proc sgplot?

Do you use proc sgplot or GTL (Graph Template Language) procedures to create your graphics?

For this course, you will be introduced to proc sgplot. There are two reasons for this.

1. When you’re interested in AS graphics, the first step is very high. There are a lot of new concepts to grasp. Having first-hand experience with proc sgplot allows you to approach GTL syntax more calmly.
2. proc sgplot already offers many graphics possibilities for professional use.

Why starting with the bar chart?

To take you a step further, you will get started with the bar chart.

There are good reasons for that:

Creating Your Graphs using a 7-Step Process

You will walk through a 7-step process for creating and customizing graphs.

After creating a first, statistically accurate version of the graph and saving it as an image in step 1, you’ll refine this graph with the remaining six steps.

Sessions with 4 Types of Activities

Each session is structured around four types of activities

1. Playground: The learner executes pieces of code to visualize the impact of certain options. This is the discovery phase.
2. Course: The trainer introduces learners to a formalized, summarized and schematized version of what was just a one-off experience.
3. Project go002: Learners make progress on their project. In addition to the course, learners have access to the Builder, a series of SAS programs consisting of code blocks to be inserted into their program and adapted to their needs.
4. Quiz: At the end of each session, the quiz helps to emphasize the importance of certain points covered that may have been overlooked previously.

Pre-Requisites

1. Be Familiar With Your Programming Environment

If you choose to use SAS® Ondemand and if you are new to SAS® Studio, you will have access to the resources “How to quickly get started with SAS® Studio” to get used to the main features needed to create folders, upload/dowload files, reset a session, and so on.

2. Manage SAS Concepts Used in the GoGoGraph Programs

SAS users can have different background when joining the course.

I have identified five SAS programming concepts used in GoGoGraph programs.

1. Dictionaries as alternative to proc contents
2. Deleting user specific global macro variables
3. Dealing with title style
4. Dealing with line breaks
5. Creating and using given formats.

One-week long, you receive one exercise per day from Monday to Friday in your inbox (MailLib Technology) to revise those five points.

3. Onboarding Session

To get started, an online session is organized.

During this session, you create the directories for the training, upload the files provided and adapt the paths in the SAS programs to your environment.

Connected to Zoom, you experiment with screen sharing, break rooms and other tool features.

At the end of this onboarding session, you take a quiz after running an initial SAS program and creating their first graph. You discover the program sections in a fun and proactive way.

Part 1 – 7 Steps to Customizing Your Graphs

Topics

Step #1

Create a first version of your graph.

Save the graph as image.

Step #2

Create your own style template to customize your graph.

Step #3

Add some text, symbol, images near the bars.

Customize the bars

• fill color, fill transparency
• pattern type, pattern color
• outline color, thickness and pattern type.

Step #4

Customize the axis.

Step #5

Use the axis tables to add statistics near the graph.

Step #6

Customize the legend.

Step #7

Use annotations (sganno=).

7-Step Discovering

Illustrated explainations

The 7 steps are detailed in a Powerpoint. Each step has about 30/40 slides.

Playground

Let’s run some code to understand how a given option behave. Programs available in the playground illustrate the 7 steps.

Quiz

Make sure you’ve spotted specific information

7 Steps in Action

Project Go002 – Instruction

Go002 is a project

Builder

Keep building

Portfolio

Part 2 – 20 guided exercises in your inbox

Reinforce your skills through practice using MailLib Technology.

MailLib offers you the opportunity to combine your professional imperatives with your desire to continue practicing and progressing.

Starts with some revisions

After discovering the 7 steps, you may want to, exercises to remind learners of a few key concepts are offered the following week.

This is a smooth transition to three weeks of guided exercises.

Save Time by Experiencing Special Scenarios

To prepare for this course, I created a portfolio of graphics. 20 are currently available.

They are a selection of graphs from FDA, EMA, Eurostat… (see images and sources) which I reproduced in SAS using the 7 steps.

On this occasion, I spotted a number of time-consuming scenarios.

They are extracted and summarized for you in 15 guided exercises (MailLib Technology).

Over the course of these sessions, the learner reproduces each of the seven steps to build a program generating a graph. This is the go002 project.

GoGoGraph monthly community meeting

Share your experiences and discuss your current graph challenges.

FAQ

Why would Experienced Programmers need GoGoGraph Solution?

GoGoGraph is designed for SAS programmers who would like to join your statistical programming department but don’t have had the opportunity to work with graphs.

GoGoGraph is for statistical programmers who only create graphs from time to time.

GoGoGraph is for statistical programmers who use trial and error to plot their graphs.

Having no time to really investigate the graphs features, they provide the minimum required output.

GoGoGraph is for small department who have to pass on the knowledge to the younger generation.

GoGoGraph is for companies still using old graph procedures, for people who want to take advantage of the most recent procedures.

GoGoGraph is for biostatisticians who want customized graphs for their papers but have to spend far too much time to implement their vision.

What is different from the resources developed by my experienced programmers?

Usually only a few programmers are really experienced with graphs in a department for many reasons

• The first steps to learn graphs are not easy. Graph program can be intimidating. Programmers usually don’t know where to start from. GoGoGraph helps programmers overcoming this first step.
• Programmers plotting graphs from time to time use examples of graphs they experienced colleagues worked on in the past. However, they are usually not fully independant. It is usually faster for them to ask experienced graphs persons that to conduct some research, living less time to those people.
• Tallented (SDTM…) programmers would be able to transition to your statistical department more easily if they can quickly gain some graph expertise.
• Graphs is a great way to retain programmers in a company. It brings a bit of diversity on a day to day basis.

In addition, have the same method used acress the deparment facilitate the maintenance and validation.

What cannot my programmers develop a similar resource internally?

How much cost an experienced programmer per week?

How much time do you think an experienced programmer would need to create a similar resource?

Wouldn’t you prefer working on specific tasks which cannot be bought on the market?

I would like to have online sessions for my team

1h30/2 hours sessions is the maximum amoung of time people can stay focus online, ideally before 1pm.

After the onboarding sessions, 7 sessions are organised.

During those sessions participants are grouped in rooms of 2 to 4 persons to work on the quizzes and the project.

Will the participants receive a certificate?

A certificate of attendance is delivered to people attending 7-step live sessions.