# What does a data-generating process (DGP) actually mean?

I am having some trouble understanding exactly what is meant by a DGP. Let's say it is stated that "the DGP is given as $$y=a+bx+e$$ where the error term fulfills all the OLS assumptions. Does this mean

a) Given knowledge of the value $$x$$ takes one would describe their belief about the value $$y$$ takes with the probability distribution on the right hand side.

b)something that allows a causal interpretation?

edit: in Econometric Theory and Methods by Davidson and Mckinnon on page 17 it is written that "the model(1.01) can be interpreted as a model for the mean of y conditional on x." To me this seems to say that the "true" regression model means what I say in a).

On Page 114 it is written(slightly summarised) "Let us suppose the model we estimate is $$y=a+bx+e$$ but that the DGP is $$y=a+bx+cz+e$$ this will yield biased estimates if $$x$$ and z are correlated". I don't get what this is supposed to mean if the regression equation describes the mean of y conditional on x. The "underspecified" model will yield a higher(or lower) coefficient to take into account the correlation, it will however still correctly describe the expectation of $$y$$ conditional on x. Here it seems to me that they are interpreting the regression coefficients as meaning the expected change in y if the regressor is changed by one unit(in a specific instance)

• "DGP" often is used without any philosophical connotation: it refers to a mathematical model. In particular, by itself it makes no psychological assertions about beliefs. The model you wrote down has no terms in it, and no implicit assumptions, that reference time or causality.
– whuber
Jan 4, 2020 at 21:36
• by belief I am referencing the bayesian interpretation of probability as a degree of belief. I know that the interpretation a) is not causal, I edited the question to make that clearer. Jan 4, 2020 at 21:42
• The point is that is one of many possible interpretations; it's not the DGP itself. If you want to understand exactly what is meant by a DGP, consult its mathematical definition or look at actual applications (regardless of the philosophical framework).
– whuber
Jan 4, 2020 at 21:45
• I don't think I follow. A DGP is a mathematical description of reality (in econometrics one seems to often abstract reality to a so called "true DGP"). What I am saying is that stating a DGP seems to allow ambiguity about what statement about reality is actually being made. Jan 4, 2020 at 21:56
• It seems like maybe you have DGP coming up in a specific context. If so, would you please provide a larger quote and citation? Jan 5, 2020 at 0:45

The DGP is the processes that cause data to occur as they do. They are a Platonic ideal that we do not and cannot know. Only in simulations can we define a mathematical DGP, in the real world they are hidden from us.

The aim of mathematical modelling is abstraction of the data. This means taking what we know and observe and trying to find a more generalised description of underlying reality that would allow us to make useful predictions in new situations. There is saying widely used on this site that all models are wrong but some are useful, this is the cause of that saying.

'the DGP is given as $$y=a+bx+e$$ where the error term fulfills all the OLS assumptions.' Is a cop out because the $$e$$ term encapsulates a wide array of lower order contributors to the data generation. Whatever has produced the data has a precise form, not a hand wavy error term. What we call error is just variation that we can't explain

a) Given knowledge of the value x takes one would describe their belief about the value y takes with the probability distribution on the right hand side.

That is the aim that we try to evaluate or beliefs about y based on x. However the example in isolation is misleading wrt the data generating process, what is quoted is a regression model not a DGP. I'd prefer to write something like $$y=a+bx+cU_1+...zU_i$$ where the $$e$$ term is split into a series of unknown underlying factors $$U$$ from 1 up to an indeterminate $$i$$. We then try to learn about $$y$$ by hypothesiseing $$y=a+bx+e$$ and projecting that model onto the data. We find that x is not quite enough to fit the data and after more poking around we realise that a previously unidentified factor is related, so we can replace $$U_1$$ with $$z$$ and collect new data to test the new hypothesis. If it fits better then we update our beliefs about the DGP. We keep going until we run out of ideas, it is no longer economically possible to collect data accurate enough to eliminate more $$U_i$$ terms, the model performs well enough for our needs or for a whole host of pragmatic reasons. We never stop because we have tried every possible $$U_i$$ term.

b)something that allows a causal interpretation?

This is getting deeper into extremely philosophical territory. Science is based on the premise that DGPs underpin reality and through careful thought and experimentation we can uncover that underlying reality. We use statistics to compare the outcome of the DGP with our hypothesis of what the DPG is and we look for a small $$e$$ to give us faith that we have captured a significant portion of the DGP. However because we never truly know the DGP we try to quantify the risk we are taking.

Let us suppose the model we estimate is $$y=a+bx+e$$ but that the DGP is $$y=a+bx+cz+e$$ this will yield biased estimates if $$x$$ and z are correlated". I don't get what this is supposed to mean if the regression equation describes the mean of y conditional on x. The "underspecified" model will yield a higher(or lower) coefficient to take into account the correlation, it will however still correctly describe the expectation of Y conditional on x. Here it seems to me that they are interpreting the regression coefficients as meaning the expected change in y if the regressor is changed by one unit(in a specific instance)

If the model has not been exposed to variation in $$U_i$$, in this case crystallised as $$z$$, it cannot account for the correlation. Part of the relationship between $$x$$ and $$y$$ is dependent on an unknown third factor which influences the nature of the relationship between $$x$$ and $$y$$. If the unseen $$z$$ changes it has an unpredictable effect on the x-y relationship because it has not been captured. If you are familiar with PCA or PLS or similar methods you will understand how subtle and complex correlations are. A correlation matrix is a high level summary that hides a lot of detail. PCA can unpack a single correlation matrix into several distinct underlying causes of correlated behaviour. Each PC describes a unique set of correlated behaviour. Furthermore each PC is uncorrelated with the others so knowing about one set of correlated behaviour gives you zero information on the others. You have to explicitly look at each possible correlation to account for it.

however still correctly describe the expectation of Y conditional on x

This will hold true while the underlying correlation structure applies, but if you haven't investigated the interaction of $$x$$ and $$z$$ then you don't know when it breaks down or changes. This issue is what underlies the need for verification of models in any new population or situation. A real world example of z may be unmeasured dietary factors affecting an analytic target (x) correlated to disease mortality (y). Over years dietary habits of populations change, which can change the metabolism of the analyte or the underlying physiology the analyte acts on and from there affects mortality in a different way.

@Carl provides some examples of commonly used scenarios for explaining DGP where we use very simple statistical models of probability to allow us to predict long run behaviour. However all these probability models have physics mechanisms underpinning them. Consider rolling dice, what factors may include that? I'll list a few I can think of :

1. Symmetry of the dice

2. Starting orientation

3. Direction of throw

4. Force of throw

5. Local topography (shape of surface its thrown towards)

6. Spin

7. Coefficient of friction between dice and surface

8. Roundness of edges and corners

9. Air movement

10. Temperature

The theory behind DGP is that if you could identify and accurately measure enough factors then you could predict the outcome of a single throw to within your desired precision. So let's say we build a model for dice rolling in a Las Vegas casino and we win so much we get blacklisted in every major casino (we forgot to lose enough). Now take that model and apply it to a poorly maintained drafty gambling den, will it still apply accurately enough to win more than we lose? We won't know until we test it.

• Regarding causal interpretation of a DGP in terms of case-control studies, my understanding is that by defining, for example, $E(Y)=\beta_0+\beta_1\times A + \beta_2 \times X+\beta_3 \times A \times X$ you're establishing a priori the true causal relationship between the treatment $A$ and the outcome $Y$ in the DGP. This allows you to evaluate how well different causal inference models (propensity score matching, exact matching, etc.) estimate the pre-defined treatment effect. Dec 9, 2021 at 16:37
• @RobertF my understanding of DGP would be slightly different, the equation in your comment as the model of the DGP. The study designer is part of the DGP since they are part of the process that leads to the eventual data. I would not describe it as 'establishing a priori the true causal relationship', rather you are curating the process of generating that data such that if your model is true you have an optimal chance of observing the expected effect, We can never know the true DGP, so we optimise for our best guess then check if it fits. This is the premise that falsification is based on. Dec 10, 2021 at 9:57
• if the data fails to fit we accept we did not understand the DGP well enough. It may be that our theory is wrong, or it may be a procedure was inconsistent, a measurement too noisy. If the data fits, we don't prove we know the true DGP, just that our model of the DGP was good enough in that case. As your equations states, the design will create the expected results, not the observed results. Those observations come from allowing the DGP to unfold. Dec 10, 2021 at 10:00
• The hope is that you understand the DGP well enough that your observed result aligns closely with the proposed model of the DGP (agreement with expected result). The deviations between observed and expected indicate how well the inference model fits the outcome of the true underlying DGP. and how completely it has been described Dec 10, 2021 at 10:01
• Agreed we can never know the true population DGP or treatment effect, we can only attempt to simulate a DGP that is as complex and noisy as the true DGP & determine if our model can detect our pre-determined Average Treatment Effect. Dec 10, 2021 at 14:00

A data generating process is a generic term for any process that generates data. For example, rolling dice, a Monte Carlo simulation of normal data with $$\mathcal{N}(0,1)$$, blowing confetti into the air to see how many pieces land inside a bucket as a function of time, throwing darts at an $$x$$-mark on a wall to show a 2D data cloud, dealing a poker hand from a marked deck of cards or whatever it takes to generate data.