How do you analyse a data table?

Often, when we want to understand how a process works, we are faced with a table of data like the one below. In this example, we want to see which parameters affect the pH of a swimming pool.

Fournisseur Pastille	Température eau	Temps entre 2 pastilles	pH
A	21.8	8.0	7.5
A	21.9	5.0	7.4
A	23.4	9.0	7.4
A	21.7	10.0	7.2

In other words, a table containing multiple rows, each of the rows containing information about the measurement we want to improve (in this case, pH) and the process conditions at the time of manufacture/measurement (in this case, the type of tablet supplier, the water temperature and the time since the last tablet). The difficulty in this type of exercise is knowing how to analyse the data and find out from the data how the process works.

💡TipDo not merge cells in Excel when constructing this type of table, as this generally complicates the analysis afterwards.

Step 1 for analysing your data: Look the table straight in the eye

Our first reaction to this type of table is often to look it straight in the eye and try to understand from the figures how the process behaves. I'll leave you to do this, but it's not easy because our brains aren't designed for it. In fact, 70% of our nerve cells are designed to analyse visual information. But a table of figures does not contain much visual information, especially figures that are difficult for us to interpret. So if you want to form an intuition, you need to move quickly on to step 2.

Step 2 for analysing your data: Building graphs

Our brains are visual, so we're going to satisfy them by giving them graphics to look at.

The graph above shows the pH measurements according to the two suppliers A and B. Immediately, it works better, we can see that the tablet supplier seems to have an influence on the pH of our pool... However, making graphs should not be the final stage in our analysis. Although it allows our brain to form an intuition, it does not provide proof that the tablet supplier does indeed have an influence on the pH of the pool. To do this, we need to statistical evidence.

Step 3 for analysing your data : Proof through statistical tests

Once you know which graphics are interesting, all you have to do is ask Ellistat to provide us with statistical proof of what we are saying by clicking on the "statistical proof" button. The following window will appear:

If you read what Ellistat tells us, you can see that the difference in averages (ANAVAR and TEST T) is statistically "very significantly different". This is proof of the supplier's influence. This is no longer intuition but proof.

Step 4 for analysing your data: Modelling the process

Perfect, we've shown the influence of the tablet supplier. That's good, but can we go further? Well, yes, by trying to model the process. Instead of analysing the columns one by one, we'll try to analyse multiple columns at once using multiple regression. The result is as follows:

Using multiple regression, not only do we see that the influence of the tablet supplier is statistically significant (see the Significance column), but we also see that the time since the last tablet is also influential. And all this in a single study - it's almost magical.

To go even further, simply go to the forecast tab and you'll be able to predict the pH of your pool according to the type of supplier, the time between two tablets and the water temperature:

So, using supplier A and a time between two tablets of 7.5 days, we predict an average pH of 7.4.

Conclusion

Analysing a data table is not as complicated as it seems. You just need to master the last 3 steps in this article to understand how your processes work. To help you in this process Ellistat guides you through the study so that you don't have to worry about statistical calculations but concentrate on understanding the physical phenomena.

Here are the modules you can use to analyse a data table:

Data Analysis

Statistics in 1 click. Use the power of statistics to find out what's behind your production data from the SPC, APC or IQC modules. Thanks to its machine learning algorithms, the Data Analysis module can be used to understand the origin of machine drift or to differentiate suppliers statistically.