This page contains the answers to various questions frequently asked about PROCESS and its creator.

**Question: Is there documentation or a user's manual for PROCESS?**

**Answer:**The current version of PROCESS is documented in Appendices A and B of the second edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysi*s. Many questions you undoubtedly will have about how to use PROCESS and what it is capable of doing and not capable of doing can be found in the documentation as well as throughout the book. The documentation is not electronically available.

**Question: How do I get PROCESS to work?**

**Answer:**The documentation answers this question, and numerous examples are found in the book. PROCESS for SPSS and SAS can be run as a syntax driven macro, and SPSS users have the option of installing a drop-down menu by installing the custom dialog file. There is a document in the zip archive containing the PROCESS files that describes how to install custom dialog files. For instructions on activating the syntax-driven macro, see the documentation.

**Question: Other than by reading the documentation or your book, are there additional places I can go to learn about how to use PROCESS?**

**Answer:**Workshops on PROCESS are scheduled in various places in the world at various times and are typically offered in the United States by Statistical Horizons or, in Europe, the Global School in Empirical Research Methods. I frequently travel, by invitation, to deliver two and three day workshops at various universities. See the "Workshops" tabs above. There is a robust collection of online discussions, YouTube videos, and so forth about the use of PROCESS, but much of it is outdated with the release of version 3.

**Question: What is the difference between version 2 and 3?**

**Answer:**There are many differences. This document outlines some but not all of the differences between the two versions. One of the more important differences is the structure of the syntax. For example, it is no longer necessary to list the variables in the model following

**vars=**, and covariates are specified using the

**cov=**option. Moderators are always

*W*or

*Z*(

*V*,

*Q*, and

*M*are never moderators in v3). PROCESS v3 also allows multicategorical independent variables and moderators in all models, and in version 3 you can program your own model rather than having to rely on models that come preprogrammed into PROCESS. Some of the preprogrammed models in version 2 were eliminated in version 3, but new models (e.g., moderated serial mediation) were added to version 3. The model templates for version 3 are different, though model numbers are consistent (e.g., model 8 in version 2 is still model 8 in version 3). Although version 3 is much more versatile than version 2, note that version 3 does not accept a dichotomous

*Y*. Eventually, version 3 will allow

*Y*to be dichotomous.

**Question: Where are the model templates for PROCESS v3?**

**Answer:**The templates for numbered models in PROCESS v3 are available in Appendix A of the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*. They are not available in electronic form except in the electronic edition of the book. The templates for version 3 are not the same as the templates for version 2. Version 3 allows you to create your own model, bypassing the model number system. For instructions on how do to this, see Appendix B of the second edition of the book.

**Question: Can I email you questions I have about the use of PROCESS?**

**Answer:**Questions about PROCESS should be directed to afhayes@processmacro.org. I try to answer questions sent to me from PROCESS users, although I will usually ignore questions emailed to me that are answered in the documentation, in

*Introduction to Mediation, Moderation, and Conditional Process Analysis*, or on this FAQ page. Thus, if I don't respond to you at all or within a week or so, there is a good chance the answer to your question is available in one of these sources. Regardless, please be patient. I receive a lot of email.

If you are going to email me a question and want to increase the probability of a response, please be very specific and tell me which version of PROCESS you are using. Questions like "Does PROCESS work for repeated measures designs?" or "It is appropriate to estimate model 1 with stacked data?" are impossible for me to answer without more details about what you are doing because the terms being used ("repeated measures" or "stacked data") are too vague to convey useful information about your design, question, or intended analysis. It is usually helpful to send me a diagram of the model you are estimating (or mention the model number) and PROCESS output, if you have one.

**Question: How do I cite PROCESS in a manuscript or publication?****Answer:**The official documentation for PROCESS is*Introduction to Mediation, Moderation, and Conditional Process Analysis*. Good academic practice is to cite something only if you have actually read it and are familiar with its content. I don't recommend using PROCESS without familiarity with what it does, as described in the book. It may not be doing what you think it is doing. I have seen many instances of researchers reporting results from the output of PROCESS that are inconsistent with what PROCESS actually is doing. These mistakes are easily avoided by reading the documentation. Many people cite a 2012 white paper I wrote that you will find circulating on line. I long ago stopped circulating this paper, and it is now outdated and not a sensible citation for PROCESS since it corresponds to version 2.**Question: Is there a way of getting SPSS to load PROCESS automatically when SPSS runs, so I don't have to manually do so each time I want to use the syntax version of PROCESS?**

**Answer:**You have two decent options. One approach for Windows users is to produce a script that will automatically load and execute each time you open SPSS. Once you have done this, you don't have to think about ever executing the macro yourself. A document to guide Windows users can be found in the zip archive that contains the PROCESS files. Thank you to Jon Peck at IBM for this recommendation. See here. However, many users have had problems getting this to work in SPSS 24. As the use of this script is not required to run PROCESS, I recommend abandoning this option you have trouble getting it to work.

The second option is to save PROCESS.sps to a particular location on your hard disk and then call it with an INSERT statement at the top of your SPSS program, before you use the macro. For example, in Windows, perhaps you have the PROCESS macro saved on your computer in a folder named "process" on the "c" drive. In that case, at the top of your SPSS program, add INSERT FILE = 'c:\process\process.sps'. When you do so, SPSS will first look for PROCESS in this location and execute it before it executes anything else in your program.

Note that neither of these options installs a dialog box in the SPSS menus.

**Question: I ran the PROCESS syntax in SPSS but no dialog box appears anywhere. What have I done wrong?**

**Answer:**Running the PROCESS syntax does not produce a dialog box or install one anywhere in SPSS. Only the custom dialog builder file, once properly installed, will produce a dialog box that you can access under "Analyze"->"Regression". Installing the dialog box does not eliminate the need for you to run the PROCESS syntax file if you intend to execute PROCESS any other way than by clicking "OK" in the dialog box. Installing the PROCESS dialog menu item is not required to use PROCESS.

**Question: How do I install a custom dialog file?**

**Answer:**The answer to this question depends on the version of SPSS you are using. The procedure has remained pretty consistent until the release of SPSS24, at which point the procedure for installation of a dialog file changed. For instructions on how to install a custom dialog file in your version, consult the instructions that come with the PROCESS files you downloaded [here is a PDF] or the documentation of the version of SPSS you are using. Installation of the custom dialog is not required to use PROCESS.

**Question: How do I know which model number to use?**

**Answer:**The preprogrammed numbered models in version 3 can be found in Appendix A of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*, 2nd edition. Choose the model number that corresponds to the model you want to estimate. If you don't know which model you want to estimate, I recommend you think about the reasons why you conducted the study, the questions you were trying to answer with the data, and/or what model best corresponds to the theory or prediction you are testing. If you don't see a model that corresponds exactly to what you want to estimate, try creating your own. See Appendix B of the 2nd edition of the book for guidance.

**Question: Is PROCESS available for any program other than SPSS or SAS?**

**Answer:**I have not produced a version of PROCESS for any other program. There are some ambitious folks who have attempted to produce versions of PROCESS for R or who provide Mplus code for some of the preprogrammed models. You can find those online, but they are outdated, in that they don't have any of the new features of PROCESS v3. I do not attest to and cannot endorse the quality of these translations or their accuracy. I simply don't know enough about them to do so.

**Question: Can you provide me with some examples of published papers based on analyses conducted with PROCESS?**

**Answer:**I don't keep track of examples of the use of PROCESS. Probably the best way to find examples is by looking at papers that include a citation to

*Introduction to Mediation, Moderation, and Conditional Process Analysis*. Click here for a list generated by Google Scholar.

**Question: Can PROCESS use sampling weights?**

**Answer:**No. Each case is weighted equally in all analyses that PROCESS can conduct.

**Question: Can PROCESS do multilevel analysis (such as multilevel mediation or multilevel conditional process analysis)?**

**Answer:**No, but MLMED will. To obtain a copy of MLMED, go to Nick Rockwood's MLMED page.

**Question: Can PROCESS estimate a model that includes reciprocal causation?**

**Answer:**I address some fairly unsophisticated but easy to implement means of entertaining questions of causal order in the book. PROCESS cannot formally estimate a model that includes reciprocal causation between two variables, such as you might do with 2SLS or something similar.

**Question: Can PROCESS estimate a model that includes a latent variable with multiple indicators?**

**Answer:**For latent variable models, I recommend Mplus, for it has the ability to estimate latent variable models and parameters that are functions of model coefficients while producing bootstrap confidence intervals for these parameters without having to jump through all the hoops many other covariance structure modeling programs require. If your "latent" variable is a average of indicators and available in your data as such, then technically it isn't a latent variable; it is observed. In that case, PROCESS could be used.

**Question: Can I use PROCESS to do a mediation (or conditional process analysis) with cross-sectional data?**

**Answer:**There is a vocal minority that takes the position that you should not or, even worse, cannot do mediation analysis with correlational data, and no doubt you will encounter a critic now and then who takes this perspective. In my opinion, this position confuses the roles of data analysis, research design, and theory in causal inference. My position on the role of data analysis in causal inference is discussed in books and journal articles I have written, and it is more relaxed, empowering, and trusting of the intelligence of the scientist than the extreme "manipulationist" position. See

*Introduction to Mediation, Moderation, and Conditional Process Analysis*(second edition, chapters 1, 3, and 4) or

*Regression Analysis and Linear Models*(Chapter 6). Other places where I discuss this include Hayes and Rockwood (2017). The position I take--that inference is a product of our minds and not our mathematics (or our software!)--has nothing to do with PROCESS. It is the position I would taken even if I never conceived or created PROCESS as a data analysis tool.

All this said, it remains your responsibility to keep your brain attuned to the inferential task at hand and not be lulled into complacency when interpreting your output (from PROCESS or elsewhere). A statistically significant indirect effect is in no way a proof of causality. Make your argument, if an argument is the best you can do given the nature of your research design, but don't overstate or convey overconfidence in what your analysis is telling you about cause-effect.

**Question: My mediator/outcome is dichotomous/count/ordinal. Can PROCESS handle this?**

**Answer:**PROCESS uses ordinary least squares (OLS) regression to estimate variables on the left sides of model equations. If you would not be comfortable using OLS regression to model one or more of your variables, you should not use PROCESS for your problem. Note that PROCESS

*will accept*count or ordinal mediator or outcomes (but not dichotomous ones), but it will use OLS regression to estimate the model coefficients. If this doesn't concern you, go ahead and use PROCESS, but anticipate some criticism from some consumers of your research if you do so.

**Question: Can PROCESS estimate a model that includes serial mediation and moderation (for example, a combination of Model 6 and Model 8)?**

**Answer:**This can be done in PROCESS version 3. Models 83-92 are moderated serial mediation models, and you can program your own if none of the preprogrammed models correspond to what you want to do. See Appendices A and B of the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*.

**Question: I am interested in estimating a moderation model (PROCESS model 1) but my independent variable X (or moderator) is categorical with more than two categories. Can I use PROCESS for this?**

**Answer:**PROCESS v3 allows the focal predictor or moderator to be multicategorical for any model PROCESS can estimate. PROCESS will automate the construction of indicator, sequential, Helmert, or effect codes, or you can program your own codes. I published a tutorial on this topic which you might find helpful for making sense of the analysis and the PROCESS output. This topic is also discussed in Chapter 10 of the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*.

**Question: Can PROCESS estimate a moderated mediation model (such as models 7, 8, 14, etc.) with a multicategorical independent variable or moderator?**

**Answer:**PROCESS v3 allows the independent variable and any moderator to be multicategorical in all of the models PROCESS estimates as well as in any model you custom program. See the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*.

**Question: I would like to estimate a mediation model (model 4) but my X is a multicategorical variable rather than dichotomous or continuous. Can PROCESS do this?**

**Answer:**PROCESS v3 allows X to be multicategorical for any model. For a discussion of mediation analysis with a multicategorical independent variable, see Chapter 6 of the second edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*. You can also read about this in Hayes, A. F., & Preacher, K. J. (2014). Statistical mediation analysis with a multicategorical independent variable.

*British Journal of Mathematical and Statistical Psychology*, 67, 451-470.

**Question: My confidence intervals for indirect effects change each time I do a mediation analysis. There must be something wrong with your code.**

**Answer:**There is nothing wrong with the code. As discussed in

*Introduction to Mediation, Moderation, and Conditional Process Analysis*, bootstrap sampling is a random resampling process. The end points of a the confidence interval are determined by percentiles in the distribution of bootstrap estimates of the indirect effect. If this bothers you, use PROCESS with a custom seed for the random number generator and use this seed each time you do the analysis. See Appendix A for instructions. Alternatively, set the number of bootstrap samples to a very large number in order to minimize sampling error in the estimation of the end points of the confidence interval.

**Question:**It appears that I have evidence of an indirect effect of*X*on*Y*through a proposed mediator, but there is no evidence of an association between*X*and*Y*. Is this possible? What should I do?**Answer:**This is not only possible, but it is probably much more common than people realize. Modern thinking about mediation analysis does not impose the requirement that there be evidence of a simple association between

*X*and

*Y*in order to estimate and test hypotheses about indirect effects. See Hayes, A. F.(2009). Beyond Baron and Kenny: Statistical mediation analysis in the new millennium.

*Communication Monographs*,

*76*, 408-420. [PDF] or Hayes, A. F., & Rockwood, N. J. (2017). Regression-based mediation and moderation analysis in clinical research: Observations, recommendations, and implementation.

*Behaviour Research and Therapy*,

*98*, 39-57. [PDF]. Also see Chapter 4 of Hayes (2018) or Chapter 15 of Darlington and Hayes (2017).

**Question: When I estimate a model using PROCESS and compare to what I get just using SPSS or SAS's regression procedure, I get different results. There must be something wrong with PROCESS.**

**Answer:**There is nothing wrong with PROCESS. When the same model is estimated using the same data with the same output options, the results will be the same as what you get with SPSS or SAS's regression procedures. There are many sources of discrepancies you may notice when discrepancies exist, and they are all generated by the user, not by PROCESS. The simplest sources involve requesting options in PROCESS that SPSS or SAS won't do on its own. A common one is requesting HC3 standard errors in PROCESS, which are different than standard OLS standard errors. SPSS and SAS won't generate these standard errors, but PROCESS will (as will my RLM and HCREG macros) but only if

__you__ask for them. When you do, standard errors, t-values, p-values, and confidence intervals are different, as they should be.

Most other sources of discrepancies are due to the user not acknowledging the existence of missing data. For example, if you mean center or standardize "univariately" (i.e., one variable at a time) prior to conducting an analysis, you will end up with variables in the analysis that are no longer mean centered or standardized after missing data are kicked out by PROCESS or SPSS or SAS's regression routine. I don't recommend doing centering or standardization computations manually. If you do, do them to a high degree of precision (three or four decimal places generally is not sufficient) and only after purging the data of cases missing on variables that will end up in the analysis. In a PROCESS model that includes moderation, PROCESS will center for you if you ask it to, and it will do it correctly (see the documentation. Also read my debunking of the mean centering myth in

*Introduction to Mediation, Moderation, and Conditional Process Analysis*).

In a mediation analysis, another common mistake I see users make is estimating the effect of

*X*on

*M*and the effect of

*M*on

*Y*controlling for

*X*in separate regressions without acknowledging the existence of missing data. Suppose, for example, some cases are missing on

*Y*. In such a situation, your estimation of the effect of

*X*on

*M*will be based on more data than what PROCESS uses, because PROCESS would discard cases missing on

*Y*before it estimates the effect of

*X*on

*M*. Although we can debate the merits and faults of listwise deletion, it is generally not good practice to piece together a mediation analysis using different subsets of the data for the estimation of different parts of the model.

Before asking for advice or bringing a "bug" in PROCESS to my attention, please check the residual degrees of freedom for the model in output produced by PROCESS (this shows up as "df2" in the PROCESS model summary section of the output) and compare it to the residual degrees of freedom from SPSS or SAS's regression routine output. If there is a difference between these, you have missing data you are not properly acknowledging somewhere. If there is no difference, then the source of the discrepancy is something else you have done differently compared to what PROCESS is doing.

**Question:**

**Can PROCESS do the kind of within-subject mediation analysis described in Judd, Kenny, and McClelland (2001,**

*Psychological Methods*)?**Answer:**See Montoya, A. K., & Hayes, A. F. (2017). Two-condition within-participant statistical mediation analysis: A path-analytic framework.

*Psychological Methods*. [PDF]. In this paper we discuss the estimation of the indirect effect and inference using bootstrapping and Monte Carlo confidence intervals. This paper also discusses parallel and serial multiple mediator versions of this model not originally addressed by Judd et al. It also discusses implementation in PROCESS v2.16. MEMORE is a macro for SPSS and SAS that Amanda Montoya designed for this kind of analysis that is a bit easier to use than PROCESS. Because MEMORE now exists, I have not and do not intend to implement this kind of analysis in PROCESS v3.

**Question:**In my mediation analysis examining the direct and indirect effects of X on Y through M, the path from X to M (or the path from M to Y) is not statistically significant. Does this mean there is no way that M could mediate the relationship between X and Y. According to Baron and Kenny (1986), it cannot. Should I bother estimating the indirect effect in this case?**Answer:**The "criteria to establish mediation" approach popularized by Baron and Kenny (1986) is historically important but not consistent with modern practice. These days, we don't rely on statistical significance criteria as described in Baron and Kenny (1986) for the individual paths in a mediation model in order to assess whether a variable M functions as a mediator of the relationship between

*X*and

*Y*. The pattern of significance or nonsignificance for individual paths in a mediation model is not pertinent to whether the indirect effect is significant. You absolutely should estimate the indirect effect. See Hayes (2009) for a brief discussion [PDF], as well as Hayes and Rockwood (2017) [PDF], Hayes (2018), and Chapter 15 of Darlington and Hayes (2017).

**Question:**I am interested in mediated moderation rather than moderated mediation. Do you have a macro for that?**Answer:**As I discuss in my book on mediation analysis, in my opinion, mediated moderation is rarely very interesting or substantively interpretable. The same model can be conceptualized in terms of moderated mediation, and the results usually are more meaningful when you change your interpretative focus from the indirect effect of a product to the conditional indirect effects. I recommend avoiding use of the term "mediated moderation" or any attempt to muster support for such a process. See Hayes (2018). Although PROCESS can be used to construct the indirect effect of a product in a "mediated moderation" model, it turns out that this is equivalent to the index of moderated mediation, and moderation of mediation is much more interesting and substantively meaningful.

**Question:**

**What is the "index of moderated mediation" I see in the PROCESS output?**

**Answer:**A discussion of the index of moderated mediation can be found in Hayes (2015,

*Multivariate Behavioral Research*). The index is also discussed in the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*.

**Question:**

**What is the "index of partial/conditional/moderated moderated mediation" I see in the PROCESS output? I don't see a discussion of this in your book.**

**Answer:**These indices and their uses and interpretation for some models are described in a Hayes (2018,

*Communication Monographs*).

**Question:**

**I estimated a conditional process model with more than one moderator. I don't see an index of moderated mediation in the PROCESS output. How can I tell if an indirect effect is moderated in these models.**

**Answer:**For a discussion of various tests of moderated mediation in models with more than one moderator, see a Hayes (2018,

*Communication Monographs*). These tests are implemented in PROCESS as of version 2.16 and also in PROCESS v3.

**Question: How can I tell whether I can claim full or partial mediation from the output of PROCESS?**

**Answer:**These are outdated concepts with little place in modern mediation analysis. They are based on the size and significance of the total and direct effects. All this information is in the output, but I recommend you avoid the use of these terms, or attempting to interpret your analysis based on the significance of the total and direct effects and whether the effect of

*X*becomes nonsignificant after adding the mediator to the model. For a discussion, see Hayes (2018) or Hayes and Rockwood (2017), which you can download from here.

**Question: I have missing data.**

**Can PROCESS handle imputed data or implement other forms of missing data analysis or procedures such as FIML?**

**Answer:**PROCESS requires complete data. It has no internal procedure for dealing with missing data other than listwise deletion. PROCESS does not integrate with the multiple imputation routines built into SPSS or SAS. If the data file you are analyzing is tagged as derived from the multiple imputation routine, it will not analyze it and an error is likely to result (The problem in SPSS is that the MATRIX language does not honor split file designations). Before using PROCESS, impute all you want, but PROCESS expects complete data and if you don't conform, it will make your data complete before analyzing it by throwing out cases missing on any of the variables in the model.

You can read about the bootstrap with multiple imputation in mediation analysis here.

**Question: The Johnson-Neyman technique is neat, but PROCESS doesn't produce regions of significance in model 1 when X is multicategorical. Why not?**

**Answer:**The mathematics for the derivation of the regions of significance are quite complicated and even impossible with more than a few groups. So PROCESS doesn't produce JN results when X is a multicategorical and specified as such using the mcx option. But check out a macro written by Amanda Montoya called OGRS that will find the boundary points for regions of significance using an iterative approach rather than a purely analytical one. OGRS is illustrated in a Hayes and Montoya (2017).

**Question: Why doesn't PROCESS have an implementation of the Johnson-Neyman method for regions of significance of the indirect effect like is implemented in MODMED?**

**Answer:**The derivations for the JN regions of significance for a conditional indirect effect discussed in Preacher, Rucker, and Hayes (2007,

*Multivariate Behavioral Research*) assume the sampling distribution of the conditional indirect effect is normal. This is a faulty assumption, and the reason bootstrapping or some other method is preferred for inference about an indirect effect. For this reason, this method is not implemented in PROCESS, and I don't recommend using MODMED for this purpose. MODMED is now obsolete, given that PROCESS can do everything MODMED can do and much more (except for this!).

**Question: Will the SAS version of PROCESS recognize a "class" command for dealing with categorical variables?**

**Answer:**No. The only options PROCESS recognizes are found in the documentation for PROCESS . See Appendices A and B of the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*.

**Question:**

**An editor/reviewer insists I have to use a structural equation modeling program instead of PROCESS. How do I respond?**

**Answer:**I address some of the differences between PROCESS and SEM in

*Introduction to Mediation, Moderation, and Conditional Process Analysis*and also in Hayes, Montoya, and Rockwood (2017). There are advantages to using SEM, but some disadvantages as well. The hard line position your reviewer or editor is taking is probably not consistent with his or her own behavior. Any OLS regression analysis is subject to the weaknesses discussed in Chapter 5 and Hayes, Montoya, and Rockwood (2017), including bias in estimation of effects due to ignoring measurement error. Yet no doubt your critics have probably used OLS regression and have probably published their own work using it, with all its flaws. And the editor has probably accepted papers with regression analyses in them. Thus, to categorically reject the legitimacy of a mediation, moderation, or conditional process analysis because an SEM program wasn't used is, at a minimum, hypocritical if not also overly ideological.

You will find some who say that it is better to use an SEM program that it is to use PROCESS. I articulate my position on this in Hayes, Montoya, and Rockwood (2017).

**Question: I have covariates in a mediation model/conditional process model, but I don't want all of them in each of the equations. How can I tell PROCESS this?**

**Answer:**In PROCESS v3, use the

**cmatrix**option as discussed in the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*. In version 2, you cannot tell PROCESS that some of the covariates should go into the model(s) of

*M*and others in the model of

*Y*.

**Question: How do I tell PROCESS which group I want to use as the reference category when using indicator coding?**

**Answer:**You can't. As discussed in the documentation, PROCESS treats the group with the numerically smallest number on the multicategorical variable coding groups as the reference. If you want a different group as the reference, recode your multicategorical variable so that your desired reference group has the numerically smallest code prior to running PROCESS.

**Question: I believe**

*X*'s effect on*Y*is moderated by two variables. How do I choose between model 2 and model 3?**Answer:**Both model 2 and model 3 allow X's effect to depend on both W and Z. Furthermore, for both models, it is possible for X's effect on Y to be statistically significant only for some combinations of W and Z. But there is an important constraint built into model 2 that does not exist in model 3. You would use model 2 if you want (or predict or hypothesize) the moderation of the effect of X on Y by W to be independent of Z. That is, model 2 would be appropriate if you want the amount by which the effect of X on Y changes as W changes to be the same across values of Z. But if you feel that the moderation of X's effect on Y by W would or should depend on Z, then model 3 is appropriate. Moderation by Z of the moderation by W of the effect of X on Y is "moderated moderation" or "three-way interaction," and this is set up and tested using model 3, not model 2. For a discussion of the mathematical distinction between these two models, see chapter 9 of the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*.

**Question:**My advisor tells me I should use the Baron and Kenny strategy for assessing mediation. But my reading of the literature tells me this isn’t recommended these days. What should I do?**Answer:**You have counted on your advisor for guidance and support during your education. Now return the favor. All but the most stubborn of advisors are open to new ideas, and many are too busy to stay informed on recent developments. Give him or her a copy of the relevant literature or a copy of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*and make your case. Also see Chapter 15 of Darlington and Hayes (2017) or Hayes and Rockwood (2017).

**Question:**Will PROCESS produce standardized coefficients?**Answer:**The regression/path coefficients that PROCESS produces are in unstandardized form. For mediation analysis, I recommend not reporting standardized coefficients when your causal agent/independent variable is dichotomous, for standardized coefficients for dichotomous variables generally have no useful substantive interpretation (see Hayes, 2018, and Darlington and Hayes, 2017, for a discussion). You can standardize your variables first prior to the use of the PROCESS, and this will generate standardized coefficients. However, the bootstrap confidence intervals you will get from PROCESS should not be interpreted as confidence intervals for the standardized effects, for that is not what they are. If you want a proper confidence interval for a standardized indirect effect, use the

**effsize**option. See the documentation. Also, be very careful when you standardized variables manually. PROCESS will eject cases from the data using listwise deletion. Make sure that before you standardize, you throw out all cases from the data that PROCESS will throw out due to missing data. If you don't do this first, then the variables you give to PROCESS after manual standardization will not actually be standardized variables. For a discussion, see the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*.

**Question:**

**I have no theoretical basis for believing there is a direct effect of**

*X*.**Is it possible to fix the direct effect in a mediation model estimated with PROCESS to zero? Or in a serial multiple mediator model, can I constrain some of the paths to zero, such as from one of the mediators to Y?**

**Answer:**PROCESS version 3 allows you to impose some zero constraints in a model, such as on a direct effect. See Appendix B of the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*. But do so at your own risk. See my position on this in Chapter 15 of Darlington and Hayes (2017).

**Question: I am getting an error message in PROCESS v3.0 that reads:**

>Error # 12410

>Source operand non-symmetric for EVAL.

>Execution of this command stops.

>Error encountered in source line # 9224

>Error # 12410

>Source operand non-symmetric for EVAL.

>Execution of this command stops.

**Answer:**This error message occurs on older versions of SPSS. Update to version 22 or newer and it should go away. I don't know the source of the error. You'd have to ask the programmers at IBM. There is no way that the the EVAL function could ever receive a non-symmetric matrix the way that PROCESS is programmed, so this error message doesn't make any sense to me.

**Question: Why was kappa-squared eliminated from PROCESS?**

**Answer:**Wen and Fan (2015,

*Psychological Methods*) showed that the derivation of the maximum possible indirect effect described in the article that introduced kappa-squared (Preacher and Kelley, 2011,

*Psychological Methods*) contains a mathematical error. As the computations in Preacher and Kelley (2011) were used to code kappa-squared in PROCESS, it seemed prudent to eliminate kappa-squared from PROCESS until this problem is fixed. Some have asked me "Then how can I get kappa-squared?" The answer is "You shouldn't use kappa-squared, so how to get it is moot."

In version 3, I also disabled the production of the ratio of the indirect effect to the total effect, the ratio of the indirect to the direct effect, and the R-square measure described by Fairchild et al. For an explanation why, see the 2nd edition of

*Introduction to Mediation, Moderation, and Conditional Process Analysis*.

**Question: Are bootstrap confidence intervals in PROCESS output produced using the percentile, bias corrected, or bias corrected and accelerated method?**

**Answer:**PROCESS version 3 produces bootstrap confidence intervals using the percentile method. This is the only method available in version 3. In version 2, PROCESS defaults to bias corrected bootstrap confidence intervals. Percentile confidence intervals can be produced using the

**percent**option in version 2. PROCESS has

*never*produced bias corrected and accelerated bootstrap confidence intervals. If you report confidence intervals coming from PROCESS using the notation "BCa" or using the term "bias corrected and accelerated" you will be misleading your audience about what you have done.

**Question:**Where did you learn to program macros?**Answer:**I am self taught. I learned to program first in BASIC on a Commodore VIC-20 my dad bought me in high school in the 1980s. My first publication was a computer program published in COMPUTE! magazine when I was 14, and for a brief period in high school I had a small software company I operated with a friend of mine that produced games and educational programs for Commodore machines. Once you learn the essence of computer programming, the skill generalizes to almost any language. The macro functions in SAS and SPSS are quite versatile, and the MATRIX language (SPSS) and PROC IML (SAS) are very powerful and can be used to program these statistical packages to do a whole lot more than what they provide to the user "off the shelf."

**Question:**Are you available for hire?**Answer:**If you are interested in learning about PROCESS or the fundamentals of mediation, moderation, and conditional process analysis, you can take a course from me in the United States offered by Statistical Horizons or, in Europe, the Global School in Empirical Research Methods. I also deliver workshops by invitation on the use of PROCESS to various organizations and universities. I generally don't do consulting on grant work, although I make occasional exceptions for projects that promise access to interesting data that I can use in my research, writing, and teaching. Otherwise, I am always interested in hearing about promising professional opportunities.