Por Borja Arrizabalaga
Incertidumbre de medida / Incertidumbre: Parámetro no negativo que caracteriza la dispersión de los valores atribuidos a un mensurando, a partir de la información que se utiliza.
Ejemplo Practico:

- Dijimos que el valor verdadero de la longitud de la pieza es de 10 mm. Esto significa que la longitud es 10,00… e infinitos ceros en la parte decimal. Ahora bien, si nuestro calibre sólo nos muestra dos decimales (hasta la centésima de mm) ¿cómo sabemos que los decimales que no vemos son cero?. Aquí ya perdemos parte de la valiosa información. Aparece una de las fuentes más importantes de duda de nuestra medida, la incertidumbre por resolución.
- Por otro lado, el hecho de retirar el calibre y volverlo a colocar entre sucesivas mediciones hace que no siempre lo coloquemos de igual manera, haciendo la misma fuerza, con el mismo ángulo de apoyo del calibre sobre las caras de la pieza, etc. Por más formación que posea el operario, no siempre logrará tomar la medida de la misma forma.
- Es bien conocido que la temperatura posee una influencia importante sobre la longitud. El valor verdadero de la longitud de 10mm es válido sólo a las temperaturas y presión atmosférica indicadas. Si, por ejemplo, la temperatura ambiente al realizar la medición es superior, la pieza se dilata y obtendremos una longitud mayor.
Cuantificacion de la incertidumbre:
- Para cada tipo de medición que realicemos debemos identificar todas las fuentes posibles de incertidumbre.
- Luego debemos identificar si son del Tipo A o del Tipo B, para poder conocer su posterior tratamiento.
- A continuación, tenemos que expresar cuantitativamente dichas componentes en las mismas unidades de la magnitud que se está analizando para que puedan ser intercomparables. Aquí es donde se realiza el mayor trabajo, ya que debemos conocer la naturaleza de cada componente y su distribución estadística.
- Por último, se realiza una tabla en donde se encuentren enumeradas todas las fuentes de incertidumbre. A esto se lo conoce como balance de incertidumbres. De aquí se obtiene una combinación de todas las componentes conocida como incertidumbre combinada. A través de un factor de confianza, y convertida en incertidumbre expandida, es la que termina acompañando al valor de medida.
Guias:
Uncertainty is perhaps the most important concept and, at the same time, the most difficult to assimilate and interpret of Metrology. It requires a broad knowledge of the measurement being carried out, the methods, the magnitude being analyzed and the magnitudes or phenomena that influence the quality of the measurement.

First let’s look at the definition that the VIM gives us:
Measurement uncertainty / Uncertainty: Non-negative parameter that characterizes the dispersion of the values attributed to a measurand, from the information that is used.
What does this definition really refer to? Uncertainty, in other words, is a parameter that quantitatively represents the «doubt» we have about the measurement performed. For a practical matter we have limited information about what we are measuring and all the sources that can influence it and how we interpret the data obtained. The values obtained in a measurement have a component of doubt, related to how reliable the measurement we are performing.
Let us remember that there is a concept, something abstract to interpret: the true value. The true value is only the actual value of the magnitude being measured. It is a value that we will never really know. We can approach it, narrow it down, but do not know exactly its value. When we get the value of a magnitude, through measurement, we have incomplete information about it. In addition to the value obtained, we must have quantitative information that indicates the quality of the measurement. Here the concept of uncertainty appears. Uncertainty involves all factors that affect the reliability of the measure. The dispersion that mentions the definition of the VIM corresponds to the variability of the values obtained in successive measures of the same magnitude. This variability stems from the fact that the methods are not perfect, or we are not looking at all the factors that influence how we obtain the measurement values and / or how we interpret and process them.
Do we see it with an example?
The best way to understand what it is about is always through an example. Suppose we want to measure the length of a metal part with a digital caliper. Our procedure requires us to take 5 successive measurements of the piece, removing and repositioning the caliber. Suppose, moreover, that it is a perfect piece of length 20 mm at 25 ° C and 1015 hPa of atmospheric pressure.
The successive measurements showed the following values:
20.01
20.00
20.02
19.99
20.01
Why does this variation exist? Why should we get different values if the piece is the same and the instrument is the same? There are many factors included here. Some important considerations:
We said that the true value of the length of the piece is 10 mm. This means that the length is 10.00 … and infinite zeros in the decimal part. Now, if our caliber only shows us two decimal places (to the hundredth of a millimeter) how do we know that the decimals we do not see are zero? Here we lose some of the valuable information. One of the most important sources of doubt of our measure appears, uncertainty by resolution.
On the other hand, the fact of removing the caliber and putting it back between successive measurements means that we do not always place it in the same way, making the same force, with the same angle of support of the caliber on the faces of the piece, etc. No matter how much training the operator has, he will not always be able to take the measure in the same way.
It is well known that the temperature has a significant influence on the length. The true value of the length of 10mm is valid only at the indicated temperatures and atmospheric pressure. If, for example, the ambient temperature at the measurement is higher, the piece expands and we get a longer length.
Those are just some sources of uncertainty, but there are more. That is, each type of measure (each method) possesses numerous components of uncertainty. If its influence is considerable, not negligible, it must be quantified.
We have in general two types of components of uncertainty:
Type A: are related to the statistical dispersion of the successive measures. The distribution can be characterized by typical deviations. Randomness plays a fundamental role in this type of components. Here are the concepts of repeatability and reproducibility that we will see later in another publication. It has to do with the real practical impossibility of performing successive measurements under exact conditions: the operator, the environmental conditions, the interpretation of the result.
Type B: are those obtained from the additional information we have. For example, a drift, a coefficient of variation of magnitude with respect to another, knowledge of the class or accuracy of the instruments
Quantification of uncertainty:
First, it is important to clarify that the correct way of expressing the result of a measurement is always by means of a measurement value (the obtained or read) accompanied by an uncertainty value that encompasses all these components.
Measurement result = Measured value ± Measurement uncertainty
The uncertainty that is expressed is a weighted combination of all the uncertainty components detected and expressed numerically.
Steps to follow:
For each type of measurement we perform we must identify all possible sources of uncertainty.
Then we must identify if they are Type A or Type B, to know their subsequent treatment.
Next, we have to quantitatively express those components in the same units of the magnitude being analyzed so that they can be intercomparable. This is where the most work is done, since we must know the nature of each component and its statistical distribution.
Finally, a table is made that lists all sources of uncertainty. This is known as the balance of uncertainties. Hence a combination of all components known as combined uncertainty is obtained. Through a factor of confidence, and converted into expanded uncertainty, is the one that ends up accompanying the measurement value.
Guides:
As is to be expected, there are many methods and magnitudes to be analyzed. It would be impossible to generalize the components of uncertainty and their analysis in a single way. However, a few years ago it was decided to draft a Guide for the Expression of Uncertainty. Known popularly by its abbreviation in English, the GUM is the document par excellence that guides us in the way to determine the sources of uncertainty and how to treat them. In addition to a statistical theoretical base, it has many examples that may resemble what we are looking for and give us a notion or an aid in determining the sources we need for our analysis. This document is freely accessible and, like the VIM, has as its main function the standardization and dissemination of good metrological practices. It can be freely downloaded in Spanish in the following link of the CEM (Centro Español de Metrología)[:]

