https://statistics.laerd.com/spss-tutorials/dichotomous-moderator-analysis-using-spss-statistics.php

https://stats.idre.ucla.edu/spss/dae/probit-regression/

Dear Ineke,

I recommend you to use PLS because you have 2 moderators. You are right that Process is not suitable for this analysis.

PLS allows you to analyze such models much more profoundly than doing regressions in SPSS.

the forum below may also be helpful for you:

https://forum.smartpls.com/viewtopic.php?t=16599

Further reading:

https://www.smartpls.com/documentation/algorithms-and-techniques/moderation

I hope this helps.

This appears to be a logistic regression because of your binary DV. Your model would appear to be logit(p)= a + b*IV + c*IV*mod1 +d*IV*mod2 + e*mod1 +f*mod2 from your description but it depends on your research question. . See: https://www.bing.com/search?FORM=U528DF&PC=U528&q=logistic+regression+spss or other stat package and my post at: https://www.researchgate.net/post/Moderation_analysis_Regression_and_categorical_variables for the rest of the methodology. Best, David Booth PS I chose logit rather than probit because your study does not appear to be a designed experiment in the sense of a full factorial experiment. see: https://www.bing.com/search?q=Montgomery+D+C+Design+and+Analysis+of+Experiments&qs=n&form=QBRE&sp=-1&pq=montgomery+d+c+design+and+analysis+of+experiments&sc=1-49&sk=&cvid=5CF71846C9414BE7A75DABA44C763974 If I am incorrect replace my logit with probit, see: https://www.bing.com/search?q=probit+regression+difference+from+logistic+regression&qs=n&form=QBRE&sp=-1&pq=probit+regression+difference+from+logistic+regression&sc=1-53&sk=&cvid=687A3FFA38344CA1BDE0888C546683A8 and :https://www.bing.com/search?q=Probit+regression+in+SPSS&pc=U528&form=QBLH&sp=-1&ghc=1&pq=probit+regression+in+spss&sc=2-25&qs=n&sk=&cvid=FC0F403CF90E445694BB7244B9335887

Thanks David Eugene Booth and Pouya Zargar your answers are very helpful! Can one of you explain me which method is better to use? A PLS or Logitistic regression? What are the advantages of using PLS over logistic regression (and vice versa)? It seems quite similar.

Thanks Ineke van Putte for important question .

Ineke van Putte The advantage of logistic regression is that it was designed to deal with your data. PLS was designed for a different type of data. See the links:

https://www.bing.com/search?q=when+should+i+use+logistic+regression&qs=AS&pq=when+should+i+use+logistic&sc=1-26&cvid=E35266A136CF431FBC6A4D1CA2C74EDB&FORM=QBLH&sp=1

and:

https://www.bing.com/search?q=when+should+i+use+PLS+regression&qs=n&form=QBRE&sp=-1&pq=when+should+i+use+pls+regression&sc=0-32&sk=&cvid=2506127E03864E5FAFAEA11261CDD57B

Notice that the two methods attack different problems. Your problem is a logistic regression. Best wishes, David Booth

My dear friend David Eugene Booth is absolutely right!

use the links to understand the depth of it.

https://www.bing.com/search?q=when+should+i+use+logistic+regression&qs=AS&pq=when+should+i+use+logistic&sc=1-26&cvid=E35266A136CF431FBC6A4D1CA2C74EDB&FORM=QBLH&sp=1

I had a misconception on this matter, which now is solved. therefore, follow the instruction that David Eugene Booth provided.

thanks to him :)

Thank you both Pouya and David! :-)

Hi David Eugene Booth and Pouya Zargar ,

I have one more question. My moderation (type of framing rational/emotional) is only on the treatment group of my Iv (so 1) not on the control group of my IV (as they did not receive any email).

When I now do the logistic regression I assume I need to do effect coding, as my interaction term (treatment * framing) does not appear in my output.

How do I need to code my variables? As I suppose that that is the problem.

So, only the level treatment of the IV has either received a rational or emotional moderation.

Do I need to code the control group -1, the rational moderation 0 and emotional 1? Or? I really don't know what to do.

Thank you again & Hope to hear from you!

Best regards,

Ineke

Ineke van Putte Normally the control group would have received a placebo email. I think in your case, if I understand correctly, that you could just say and code that placebo was no treatment. I believe the coding you suggest will work. However, if you are using SPSS it may force a coding on you. If that happens, let's talk again. One more reason for avoiding SPSS. Best, David

as David Eugene Booth mentioned, this coding may work but SPSS often forces you to re-code. this is why i suggested PLS at first but you can still go through with it in SPSS. you can customize your analysis before clicking on the test.

Hi David Eugene Booth and Pouya Zargar ,

Thank you for your response! Unfortunately I am doing my analysis in SPSS, which means that I need to re-code.

Would the following work?

DV: 0 (not active) & 1 (active)

IV: -1(control) & 1(treatment)

Mod framing : -1 (Control & rational) & 1 (emotional)

Many thanks again!

Ineke van Putte I don't think your new system will work. I don't work with SPSS but you might look at this link:

https://www.bing.com/search?q=SPSS+DEFAULT+DUMMY+CODING&pc=U528&form=QBLH&sp=-1&pq=spss+default+dummy+coding&sc=1-25&qs=n&sk=&cvid=8ADBE5A3050B48098F87326C3F3D82D9

See what you need to do to get your interpretations correct. Here is a link that lead me to a much easier life than SPSS: http://sgpwe.izt.uam.mx/files/users/uami/gma/R_for_dummies.pdf The corresponding regression notes are attached. May the Force be with you. Best, D. Booth

if you try that formula, it may work .... if you insist on using SPSS carry on, but i recommend you give all codes a positive value (e.g. 1 / 2 ) this may cause reverse results which can be undesirable..

at the end you know which code is what so you can easily interpret the results.

there are many techniques to get this over with. have a look at this link, it may give you some ideas of how to go further:

https://www.wbiss.lawrenceemeagwali.net/

I'd personally do not rec-ode anything in SPSS unless it is a transformation as it may then cause other unwanted issues. if you re-select your approach you can then have same codes but test them differently. the test itself is not as important as the approach and how you define your data and variables.

then again, you have also R which is helpful in this scenario.

Nice Dear Pouya Zargar