SPSS - Moderation Analyses with Simple Slopes + Process

Поделиться
HTML-код
  • Опубликовано: 13 янв 2025

Комментарии • 402

  • @philvlachou4814
    @philvlachou4814 4 года назад +106

    00:35 G*Power
    3:23 Centring
    5:16 Missing data
    6:52 outliers
    20:10 multicollinearity
    21:42 normality, linearity, homogeneity, homoscedasticity
    22:30 setting up PROCESS
    30:48 interpret PROCESS output
    52:15 Create line chart from PROCESS output

    • @maticasamyu
      @maticasamyu 3 года назад +2

      This is really HELPFUL. Thanks!
      Honestly, I know right away if there's gonna be this time choices in the comment section. Then when I scrolled down, here it is I found in the first row, awesome

  • @dr.remicoker6551
    @dr.remicoker6551 6 лет назад +73

    This is the most amazing thing that I have ever seen. Can you please just do...everything for all of stats...ever? I learned more from this video (for my dissertation no less), than I have ever learned in 4 years of statistics!

  • @daothang7892
    @daothang7892 10 месяцев назад +1

    What an amazing video ! Your teaching is so useful, practical and easily understandable. Thank you so much !!!

  • @xllvr
    @xllvr 5 лет назад +5

    Currently doing an assignment in school and the professor didn't really explain how to report a moderation analysis. Watching this video was significantly better than both class and the hours of research that I spent trying to figure things out. So thanks very much; your channel has been and remains to be extremely educational in many of my research endeavors.

  • @richardjones6953
    @richardjones6953 5 лет назад +1

    Dr. Buchanan, this is just a little note to express my gratitude for all your excellent tutorials! They are hugely appreciated. Thank you!

  • @JanineAlexander-i4j
    @JanineAlexander-i4j 2 месяца назад

    Dr Buchanan thank you so much for your work and videos. The detail you provide, the explanations you share are priceless. You are incomparable, and a born and gifted teacher. You help us to really see and really understand. Like the others below you have made such a big big difference to my stat studies and what I can now achieve. Thank you so so much.

  • @1duta
    @1duta 9 лет назад

    This video is awesome. Great narration and video quality! It is truly one of the best educational videos that I have ever seen at RUclips.

  • @elliegreyz8548
    @elliegreyz8548 11 месяцев назад +2

    Thank you! You explained something I thought I’d never understood so clearly. I think my report might actually make sense.

    • @StatisticsofDOOM
      @StatisticsofDOOM  11 месяцев назад

      Thanks for the kind words! Glad it helped.

  • @alisonpike8421
    @alisonpike8421 8 лет назад +1

    THANK YOU. I am a senior-ish academic who should know better, but was struggling to figure out how to plot simple slopes. This video is amazing -- great resource. Your generosity is much appreciated.

  • @edgarmesquita
    @edgarmesquita 6 лет назад +1

    You are an amazing lecturer. Extremely thorough, but eloquent at the same time.

  • @xiaohanhu5272
    @xiaohanhu5272 5 лет назад +1

    This is just so helpful! I am struggling with doing moderation analysis with continuous variable and did not even realize I can do all this using process macro (typically just use the macro for mediation analysis). This video is just so clear in explaining everything you need to do and how to interpret everything you get. Thank you so much for this!

  • @fnssupply
    @fnssupply 6 лет назад +1

    Thank you Dr. Buchanan, this video was very helpful. I wish stat professors actually taught like you.

  • @TofahaGirls
    @TofahaGirls 8 лет назад +1

    Thanks for this video. I was having such a hard time with moderation, especially the PROCESS plug-in. Really excited to watch your other videos!

  • @jiaqizhu4705
    @jiaqizhu4705 7 лет назад +2

    Really thanks. I got a lot from it which is the best video for moderation in process in RUclips.

  • @liuqing1995
    @liuqing1995 Год назад +1

    amazing presentation of hard-core statistics! wish I could know your channel early!

  • @FiadorBismark
    @FiadorBismark 3 года назад +1

    Excellent video, with easy to understand interpretation.

  • @rachelfield3205
    @rachelfield3205 3 года назад +11

    You've just saved my dissertation, I think I'm in love with you

  • @timothyclay1943
    @timothyclay1943 6 лет назад +1

    Thanks Dr. Erin, your explanation here was super easy to understand and helped greatly with one of my postgraduate assignments!

  • @bulgariantreasure
    @bulgariantreasure 9 лет назад

    Hi, Erin! First off - thank you so much! This video was unbelievably clear and helpful. I just installed the Process add-on and started playing around with it for my dissertation. Your video was so much more thorough than Andy Field's :) I love him, but you made it all come together with ease...I will definitely subscribe and try to watch some of your other videos. Also, going through all the assumptions first in the video was essential for people who are new to all of this. Thank you again!

  • @visibleOnlyNaNaNA
    @visibleOnlyNaNaNA 8 лет назад +3

    love your voice and your clear, step-by-step explanation!!!

  • @jastootie
    @jastootie 7 лет назад +1

    Thank you Dr. Buchanan! Your videos are helping me tremendously with my Master's thesis analyses!

  • @wandickleao4634
    @wandickleao4634 8 лет назад

    Erin, I'm very happy for this video. It help me in the analysis and comprehension of date bank. I'm concluding Master's Degree in Business Management. Thank's again. Sorry for the english I'm Brazilian and i'm development the english language.

  • @ivanal2615
    @ivanal2615 8 лет назад +1

    Thank you so much for this great video, it helped me a lot :) As a thank you I have a little tip for you: change your Word Font to Courier (the same font is used in SPSS output for process) - it will keep the all the numbers neatly under each other :)

    • @StatisticsofDOOM
      @StatisticsofDOOM  8 лет назад

      Good to know ha! I usually put things into Excel, but Word can be a bit easier to read for these examples.

  • @Msichanawana
    @Msichanawana 9 лет назад

    EXTREMELY INFORMATIVE VIDEO!!!!!
    Good work explaining what everything does in the PROCESS dialog and output boxes!
    Thank you very much Erin.

  • @lisab9951
    @lisab9951 Год назад +1

    best best best video out here.

  • @lindabouali6705
    @lindabouali6705 7 лет назад

    Hey there ! Just a piece of advice- you can select for the names to be run as long names if you go to "long names" option :)

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      That's a newer option! Sometimes it still doesn't run, so I try to go with short names as much as I can.

  • @korwinkressmann425
    @korwinkressmann425 3 года назад +1

    Thanks a lot for this video. Your teaching is great and you sound super sympathetic.

  • @TwilightFlip
    @TwilightFlip 3 года назад

    I could cry, this is so helpful for my thesis paper, thank you!

  • @jazmcmahon3434
    @jazmcmahon3434 6 лет назад +1

    This is the most informative video, thank you so much you have just saved my Honors Thesis & my soul!

  • @M3Lucky
    @M3Lucky 7 лет назад +3

    Thank you so much for this video! It is extremely helpful, not only in understanding and interpreting the whole process, but also formatting the graphs :)

  • @damlaaltay
    @damlaaltay 2 года назад +1

    You are a Godsent! Thank you so much for this video, helped me so much! ☺

  • @halabishki
    @halabishki 6 лет назад +2

    You are an honest to God actual angel. THANK YOU SO MUCH.

  • @Sherlock9830
    @Sherlock9830 5 лет назад +1

    Outstanding lecture, this has helped me enormously with my thesis

  • @muhammadatique3665
    @muhammadatique3665 5 лет назад +2

    JOB WELL DONE, that was amazing, you cannot even imagine how much you helped me, lets just say you cannot even measure in SPSS :D, Stay blessed

  • @gulafshan9388
    @gulafshan9388 7 лет назад +1

    I really like your way of making things simpler to report. Thank you

  • @treblesome
    @treblesome 8 лет назад

    This was an incredible help - THANK YOU for your clear explanations, and for taking the time to educate us all!

  • @life_atlas
    @life_atlas 6 лет назад +3

    hello i'm writing from Turkey.. Thank you for the video... ı used your way my thesis...

  • @virginiatompkins7656
    @virginiatompkins7656 2 года назад +1

    Can you have multiple independent variables in Model 1 (not covariates, but variables that interact with the moderator)? If not, is there any way to do this with PROCESS? I have four IVs and one DV and I want to test for age as a moderator of each relationship.

    • @StatisticsofDOOM
      @StatisticsofDOOM  2 года назад

      I think you can only do up to two-two way interactions if I remember correctly. You would need to do it manually if you want to do all of them together in one model.

  • @user-ox8po9bo3r
    @user-ox8po9bo3r 7 лет назад +1

    Thank you so much, you safed my live respectively my bachelor thesis!

  • @GertRijlaarsdam
    @GertRijlaarsdam 5 лет назад +2

    Marvelous, thank you, full context, with reporting advices etc.

  • @amoouuii07
    @amoouuii07 8 лет назад

    Hi! Just wanted to say THANK YOU! Video was very helpful your responses to questions in the comments section was very helpful! Thank you thank you thank you!!

  • @evachen9204
    @evachen9204 4 года назад +1

    Thank for your video! It helps me so much on my assignment!!!

  • @aldrea7
    @aldrea7 5 лет назад +1

    Thank you for making sure I didn't look like an idiot in front of my thesis committee member when we had an analysis consultation today

  • @chesskaboomgustz3755
    @chesskaboomgustz3755 5 лет назад +1

    Thanks for sharing. It helped me a lot. Keep up the good work.

  • @sofiavintimilla4852
    @sofiavintimilla4852 4 года назад +1

    I'm here to say thank you ! This really helps

  • @nachenache3223
    @nachenache3223 5 лет назад +2

    This lesson was incredibly helpful and clear. Thank you for uploading this!

  • @seanfdaly4
    @seanfdaly4 4 года назад +1

    Thank you!!! This was simple and easy to understand

    • @StatisticsofDOOM
      @StatisticsofDOOM  4 года назад

      Thanks! Appreciate the kind words.

    • @seanfdaly4
      @seanfdaly4 4 года назад

      Statistics of DOOM I do have a question...how are the levels calculated if not just by SD? I did moderation and used the percentage points of 16, 50, and 84%...is that accurate?

    • @StatisticsofDOOM
      @StatisticsofDOOM  4 года назад +1

      @@seanfdaly4 You can do it that way as well, but do realize that's basically the same idea as the SD, since those percentages correspond to 1 SD in the z normal distribution.

  • @M3Lucky
    @M3Lucky 7 лет назад +1

    Question about the model summary statistics (ie. 31:46)...since this is a moderation regression, with 'books' as the IV, and 'attendance' as the moderator; could you interpret it as: 'books' accounts for 40.22% of the variance in grades when moderated by 'attendance'? Or is this wrong?

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад +1

      Yeah something like that - for the main effect it's basically that it accounts for 40% of the variance controlling for attendance and the interaction of the two.

  • @juandiegovaamonde4683
    @juandiegovaamonde4683 9 лет назад

    Thanks Erin for this video! It was really useful.
    I was wondering if you are going to upload a video on MODERATED MEDIATION with PROCESS, especially one dealing with several IVs. This type of analysis is becoming popular in psychology papers. Thank you!

  • @lazliz7628
    @lazliz7628 6 лет назад +1

    Thank you so much for such nice explanation and going through the important steps!

  • @alyssasinger6582
    @alyssasinger6582 8 лет назад +2

    This video was so helpful, clear and understandable. Thank you so much! Do you have a video on interpreting output with multiple moderators? (Model 2)

  • @marig5587
    @marig5587 5 лет назад +1

    You are truly the best.

  • @supratikmondal3108
    @supratikmondal3108 2 года назад +1

    Thank you for the video. It was very helpful...

  • @iROdawnte
    @iROdawnte 8 месяцев назад

    Hopefully someone reads this but at 34:12, why is that predictor significant with 0.148? Is it because it's less than the constant (a) of 61.6024? Thanks!

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 месяцев назад

      It is significant if you use p < .05 because the p value is .0148 (you are missing the decimal placement).

    • @iROdawnte
      @iROdawnte 7 месяцев назад

      @@StatisticsofDOOM Thank you, it was a silly question lol

  • @nurankuzu4438
    @nurankuzu4438 3 года назад +1

    Thank you so much!!! Very helpful 🙏🏻

  • @khizarhayat1816
    @khizarhayat1816 3 года назад +1

    It's amazing, I learned a lot from this lecture. Thanks for help!

  • @ewajanas9497
    @ewajanas9497 4 года назад +1

    Thank you so much!! This video is amazing, explained everything I needed to know and more!

  • @nazzy_knows
    @nazzy_knows 6 лет назад +1

    Thank you so much for this video. It really helped fill in that gaps that I really needed to know to finish my research write up :)

  • @inexpensivetrick
    @inexpensivetrick 9 лет назад

    A prayer answered! This is so incredibly helpful! Thank you, thank you, thank you!!!

  • @sabrafarrukh9857
    @sabrafarrukh9857 7 лет назад +3

    Thank you so much. You have explained things in a wonderful way.

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад +2

      Glad to be of help!

    • @sabrafarrukh9857
      @sabrafarrukh9857 7 лет назад

      I have applied same things which you explained in video. My interaction is Insignificant ( in video you said there is no specific explanation of significance and gave Three level L , M , H. Please guide me literature reference from where you took this explanation.) My supervisor wants to see literature reference as well as book reference. Please give me one more favour that provide reference of explanation of Johnson-Neyman Techniques. If you have used all these in any of your publication or any other. Please accept my advance Thanks :)

  • @gigikullawadee2023
    @gigikullawadee2023 5 лет назад +1

    You'll surely be a part of my graduation speech.

  • @carimarvelli
    @carimarvelli 7 лет назад

    Hi Dr. Buchanan. Thank you for breaking that down so simply. Really helped me understand how to interpret the interaction.
    One question: How would you interpret the output for a dichotomous moderator? Your data has 3 levels of attendance with low, medium and high, where mine has only two; one is the control condition (0) and the other the experimental (1). I am looking at pre- and post-test scores, where the pretest scores are my predictor variable, post-test scores the DV, and the condition(s) the moderator.
    My output did not include the Johnson-Neyman method because it could not be used with a dichotomous moderator, but it did note that the moderator was mean centered. How does it mean center a dichotomous variable?
    Also, what if you were to include covariates?
    Thanks again!
    Cari

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад +1

      I would turn off mean centering if you use a dichotomous moderator. My variables are continuous, so it created the low, average, and high for me. Mean centering does not make sense for categorical variables.
      If you have a dichotomous moderator, you can interpret it as the simple slope for each group separately.

    • @t1cketyboo
      @t1cketyboo 5 лет назад +1

      @@StatisticsofDOOM Hi Erin, I have a similar question - for my predictors, one is continuous and the other is dichotomous - how would I check assumptions are met for this analysis? Also - this video has helped me immensely, thank you so much!

    • @StatisticsofDOOM
      @StatisticsofDOOM  5 лет назад +1

      @@t1cketyboo if it's only dichotomous (meaning only two categories), you can do the data screening like normal and shown in this video - you may see some weird splits in the plots because of the groups, but generally if it's ok assumptions wise, that doesn't affect too much.

  • @LeVacarois
    @LeVacarois 8 лет назад +1

    Thank you very much for this detailed video, what a great resource. I was particularly confused regarding the use of GPower in this situation, and your video helped a lot. Unfortunately I am still unsure as to how to approach this for my particular situation. I am interested in the interaction between a categorical IV with 4 levels, and a continuous moderator. Does this still mean I need to enter '3' as the number of predictors (1 IV, 1 moderator, and their interaction) in GPower? Or do I need to count each dummy variable that PROCESS calculates as a predictor and thus see this as using 7 predictors (moderator, D1, D2, D3, int_1, int_2, int_3)? I assume it is the latter, but I am just not entirely sure... I am sorry to ask, but have been unable to find a clear answer to this anywhere...

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      Use each of the dummy coded variables, so 7 yes. :)

    • @LeVacarois
      @LeVacarois 7 лет назад

      I'm sorry for my late response. Thank you so much for your reply! Your channel is a very valuable resource.

  • @Jajalaatmaar
    @Jajalaatmaar 7 лет назад +1

    Hey Erin, thanks for your video. It's helpful to go through it step by step. I've run into a weird problem: doing a linear regression analysis in with 2 predictors + interaction term in spss I find p=.030 for interaction. That's why I'm figuring out PROCESS now to see how the interaction works. However, in PROCESS, I find p=.108 for the interaction. I DO however find conditional interaction effects with 2x p

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      Something must be different ... maybe try centering your variables before running regular regression? Then try the process on the non centered variables? It should not turn out that different!

  • @lauravanbethraij6723
    @lauravanbethraij6723 3 года назад

    If i have got 2 independent variables, 1 dependent variable, and 1 moderating variable, should i add the moderating variable to the independents at 7:05 ? (Yes i have no idea what i am doing)

    • @StatisticsofDOOM
      @StatisticsofDOOM  3 года назад

      Yes, but if you have two two-way interactions because you have two X variables, you will need to use Model 2.

  • @ryljaco
    @ryljaco 8 лет назад +1

    Hi Erin, what are some steps for troubleshooting bent and crooked lines in the line graph? For some reason SPSS includes 0 as a value on my X-axis!
    Thanks.

  • @koenmunk7043
    @koenmunk7043 9 лет назад +2

    Hey, I've got a question: At the 'Conditional effect of X on Y at values of the moderator' part of the output, my p-values are all insignificant, what should I do.
    Also, there were no significant direct effects, but my interaction effect is significant (p = .0492)

  • @PopTherapy
    @PopTherapy 2 года назад

    Hi Erin, thanks for this great video! I had a quick question about determining the direction of moderation. If the interaction is significant (and a positive coefficient) but the conditional effects are negative numbers, and also significant, do we rely on the conditional effects when determining the direction of the moderation?

    • @StatisticsofDOOM
      @StatisticsofDOOM  2 года назад +1

      You would only look at the simple slopes to explain the effect, as the interaction coefficient has the other variable also included (and that coefficient isn't really interpretable anyway, except that it is "not zero").

    • @PopTherapy
      @PopTherapy 2 года назад +1

      @@StatisticsofDOOM amazing, thank you so much for getting back to me! I'd been so confused about this. :)

  • @JimBob-wi3rq
    @JimBob-wi3rq 7 лет назад +1

    Hi Dr. Buchanan, love your videos. They are really clear and helpful. And really liked the simple slopes in this vid, but some of the mediation analyses may be misleading.
    I've been struggling with interpreting mediation results using Process. Particularly, I think I was wrongly using the causal steps approach to mediation (i.e., significant a path, b path, c path and c' path) to interpret Process' bootstrapping results (i.e., estimating confidence intervals). However, Hayes (2014) cautions against using the casual steps approach for several reasons (assumptions of normality, weak power, "inconsistent" mediation per Baron and Kenny, etc...) and cautions against using the outdated terms of full and partial mediation. See Hayes (2014), section 6.1
    What do you think?

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      I don't disagree. There are lots of approaches to mediation. I propose presenting your results in the most understandable way possible, which would allow for others to decide if they believe the evidence you are presenting. Additionally, I would not use the word "cause" without a supporting experimental research design (which is a different issue than the question of mediation).

  • @hollydixon919
    @hollydixon919 4 года назад +1

    This was so fantastic! Thank you

  • @laurahelm7399
    @laurahelm7399 7 лет назад +1

    What does it mean when the effect is negative? So the group scoring below the average of the moderator has a negative effect?

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      Negative predictors are interpreted: as X increases, Y decreases.

  • @d.andygodfrey9857
    @d.andygodfrey9857 5 лет назад +1

    Dr. Buchanan, Thank you so much for this video. It has been extremely helpful. I am writing up the results for my moderation model and I was wondering if you had citations for the process in which you screened and removed any leveraging outliers prior to creating running the moderation. Any information would be helpful. thank you.

    • @StatisticsofDOOM
      @StatisticsofDOOM  5 лет назад

      You can use Cohen Cohen Aiken and West for leverage/cooks and Tabachnick and Fidell for the mahalanobis.

    • @d.andygodfrey9857
      @d.andygodfrey9857 5 лет назад +1

      @@StatisticsofDOOM Thank you again!

  • @tbrifki
    @tbrifki 5 лет назад +1

    Great lecture, thanks!

  • @harahara3544
    @harahara3544 8 лет назад

    Thanks Erin, great video. I need some clarification on the POWER that you touched upon earlier in the video. As far as predictors (K) are concerned, if I have one IV, one moderator, 1 mediator, and 1 DV, what is the total number of predictors? In other words, sample size required for the medium effect size and power .8. Thanks for your help...

  • @nannielane
    @nannielane 3 года назад +1

    I cannot like this video enough!!

  • @sornakan
    @sornakan 8 лет назад

    Erin, Wonderful Video. Kindly let me know how to interpret for a continuous independent variable since the video is for a categorical variable.

  • @9197samantha
    @9197samantha 7 лет назад +1

    Hi this video is extremely helpful but I have a question, what is the values in the coefficients column are negative, what do they imply? Thank you so much!

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      You would interpret that just like a negative correlation ... as X goes up Y goes down.

  • @diwang6773
    @diwang6773 7 лет назад +1

    Thank you so much for this video, which is so helpful to my essay writing.

  • @olgabiaobrzeska6130
    @olgabiaobrzeska6130 7 лет назад

    Hi Erin, when explaining how to report the interaction, you interpret coeff as b (standardized coefficients) while (correct me if I'm wrong) they are unstandardized coefficients in fact.I would like to be able to interpret the strength of the predictor (X) and also see if it is comparable to the previous findings on this topic. If you could relate to this issue, would be great. Thanks!

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      Right, so in process they are mean centered, and not z-scored. They are unstandardized, but definitely 0 centered.

  • @williegaloo
    @williegaloo 9 лет назад

    Thanks for sharing, Erin! Love your video.

  • @yaraib2720
    @yaraib2720 7 лет назад +1

    Thank you so much Dr. Buchanan, your video is excellent and very very helpful. I have a question that maybe you can help with. I ran moderation and the interaction effect was non-significant, however; in the conditional effect table (high and low groups), all effects become statistically significant. Does this mean I can use this and say moderation does occur for these specific groups?

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      No - that means you do not have significant differences between slopes, but your main effect of your variable is overall significant.

  • @aubreyjones4674
    @aubreyjones4674 5 лет назад +1

    This was so helpful, thank you! My interaction effect was not significant. Would this be a reason for not seeing the johnson neyman output or the conditional effects? I should add that I am also using the new version3.3

    • @StatisticsofDOOM
      @StatisticsofDOOM  5 лет назад

      Yes, I think you only get the Johnson Neyman when the interaction is significant. There's a button you can tell it to give you the interaction either way.

  • @sabz1023
    @sabz1023 9 лет назад

    Hey Erin! Wonderful video, extremely useful! I do have a quick question though - I have one continuous moderator and one categorical IV with four levels. I am guessing I have to dummy code the variables and then enter those into the model? Could you please throw some light on this? My DVs are continuous too. Thanks!

  • @benriverman6823
    @benriverman6823 8 лет назад +1

    I have significant X-> Y effect, Johnson-Neyman output, and for simple slopes, significant p-values for high and centered average of my mediator. However, my interaction was not significant. What does this mean, and can I still interpret my simple slopes?

    • @nishthamehra26
      @nishthamehra26 8 лет назад

      i have similar results. The independent regression coefficients are significant but the interaction is non significant. Can i interpret moderation still? as some books say that moderation is actually the interaction between the IV and Moderator.

    • @StatisticsofDOOM
      @StatisticsofDOOM  8 лет назад +2

      That means that there is main effect of X to Y, but the slopes do not different across M. I would not interpret the simple slopes because the interaction is not significant.

    • @StatisticsofDOOM
      @StatisticsofDOOM  8 лет назад +2

      I would not interpret the simple slopes if the interaction coefficient is not significant, as that means all the simple slopes are statistically equal (given power, of course) ... the interaction is between X and M predicting Y, correct, which is how it is covered in this video.

  • @simonspenterifversen3329
    @simonspenterifversen3329 3 года назад +1

    Great video! I can't get the "Conditional effect of X on Y at values of the moderator" output in version 3.5 of PROCESS? Can someone help me on how to get this so I can measure the slopes like in the video?

    • @StatisticsofDOOM
      @StatisticsofDOOM  3 года назад

      There’s an option to change the p value at which the slopes print out. Otherwise if the moderator is not below that p value it won’t print out.

  • @blainemarcano
    @blainemarcano 5 лет назад +1

    Thank you Dr Buchanan.

  • @darrenka2735
    @darrenka2735 8 лет назад

    Thanks for a great video. I have a question related to slope. Can I consider the Johnson's values as LOW and HIGH for a slope? Thanks.

  • @Rubinovitsch
    @Rubinovitsch 4 года назад +1

    1000th like! Well deserved for a very informative lecture (:

  • @maxg.6810
    @maxg.6810 9 лет назад

    Hi Erin, I have a question regarding the interaction term in the output. Do I have to include it in my write up and what does it mean if the overall moderation model is significant but the interaction term is not? Thank you very much in advance, Max.

  • @annamarini7805
    @annamarini7805 9 лет назад

    Hi Erin, Thanks for the videos. They are all very helpful! I was wondering whether you were planning to do more videos on PROCESS, especially on how to use it when you have an indip. variable with 3 categories (moderator and dependent variable continuous). Looking in the internet it seems that many people are asking the same question. Plus how do you to report PROCESS results in APA style? Any help would be much appreciate it. Thanks!

  • @xThomas79x
    @xThomas79x 8 лет назад +1

    Hi Erin!
    Thanks for all your videos that helped me a lot! But just one question: Does there actually a graphical user interface for PROCESS exist? Or should I better use SmartPLS or AMOS for this? The problem is that with these programs it is not possible to calculate with moderations.... only mediations are possible (as far as I know)! :((

    • @StatisticsofDOOM
      @StatisticsofDOOM  8 лет назад +1

      I'm sorry, I don't understand the question. PROCESS is graphical through SPSS and does both mediation and moderation.

    • @xThomas79x
      @xThomas79x 8 лет назад

      I am sorry for being imprecise, but I meant whether it is possible to "draw" / visualise your calculated model with PROCESS. Like it is in AMOS (you have the regressions of the SEM AND the graphical model with its arrows and boxes).
      Have a nice holiday.... ;-)

    • @StatisticsofDOOM
      @StatisticsofDOOM  8 лет назад

      No on the graphic interface to Process. I think you might be able to coerce AMOS into interactions, but you would have to make the interaction column manually, and then interpreting would be much harder than the regression way (also df is treated very differently).

    • @StatisticsofDOOM
      @StatisticsofDOOM  8 лет назад

      Also, sorry, I wasn't thinking the right way earlier :| thanks for the clarification!

  • @konstantinmetzger9704
    @konstantinmetzger9704 7 лет назад +1

    I really like your video. It helps me a lot. There are still some question remaining though:
    1. How do i identify non-significance? Just by looking at the p-value? Or is the LLCI and ULCI also important?
    2. What if my overall Effect is non significant, but one of my moderation effet in the slopes is significant? Am i allowed to predict a relationship between my X on Y in that case?
    I´d be glad for some help. Thank you!

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад +1

      1) you could do either depending on your area's focus on p values, I would say p values are more comment.
      2) that would depend on if the interaction was significant - in general main effects and simple slopes are not interpreted the same.

    • @konstantinmetzger9704
      @konstantinmetzger9704 7 лет назад

      Thank you Erin, your answer helped me a lot. However when analysing my last Hypothesis i stumbled across another problem:
      The moderator is gender. My interaction is non significant (p = .95), but one of my two predictors is significant (p

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад +1

      Hi again, the interaction isn't significant because those two intervals overlap ... so the main effect of that variable is significant (your p

    • @konstantinmetzger9704
      @konstantinmetzger9704 7 лет назад +1

      Thanks a lot! You've been of great help! Keep up the good work. Best regards

  • @dancohn3703
    @dancohn3703 5 лет назад

    Hello Dr. Buchanan- I will echo the sentiment of many others here and thank you deeply for sharing this video. It has been crucial in the process of proposing and defending my dissertation.
    I have one question- In the initial screen when opening up the process macro, under "options", I have a drop-down menu for "Heteroscedasticity-consistent inference", rather than a checkbox for "Heteroscedasticity-consistent SEs" like in your video. This drop down menu presents several options such as HCO, HC1, HC2, HC3, HC4, as well as "none" (some of these different HC's also have names in parentheses afterwards, I'm assuming credited to different researchers). Do you have any insight about which one of these options I should choose? Thanks!

    • @StatisticsofDOOM
      @StatisticsofDOOM  5 лет назад

      Let me refer you to Hayes' publication on this topic:
      link.springer.com/content/pdf/10.3758/BF03192961.pdf
      Specifically I find this paragraph useful:
      For small sample sizes, the standard errors from HC0 are quite biased, usually downward, and this results in overly liberal inferences in regression models (see, e.g., Bera, Suprayitno, & Premaratne, 2002; Chesher & Jewitt, 1987; Cribari-Neto, Ferrari, & Cordeiro, 2000; Cribari-Neto & Zarkos, 2001; Furno, 1996). But HC0 is a consistent estimator when the errors are heteroskedastic; that is, the bias shrinks with increasing sample size. Three alternative estimators, HC1, HC2, and HC3, are all asymptotically equivalent to HC0 but have far superior small sample properties relative to HC0 (Long & Ervin, 2000; MacKin-non & White, 1985). A newer estimator, HC4, is preferred when there are cases with high leverage.

  • @deryakaratas6693
    @deryakaratas6693 8 лет назад +2

    Hi. Thank yuou for your video. I have a problem and if you answer ı would be pleased. My p value in model summary is significant however interaction is not significant. So how can ı interpret it? I know there is no mderation but can ı say that x effects y directly from looking at model summary? Thank you.

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад +1

      You will need to see if the main effects are significant to say x predicts y, look at that predictor value.

  • @AngstyLou
    @AngstyLou 8 лет назад +1

    Very clear delineation. Thank you!

  • @niklaskuhme4235
    @niklaskuhme4235 3 года назад

    Hi, amazing video - really helped me with many of my projects so far!
    Quick question regarding the outlier "analysis": Is there a reference I can use (as in research paper I can quote) that suggests the 2/3 rule for Mahalanobis, Cooks, and Leverage? Personally, I have only found academic articles that suggest to either use only one or maybe two of them. Thanks in advance for any help! :)

    • @StatisticsofDOOM
      @StatisticsofDOOM  3 года назад

      I know people have cited my videos, but for academic sources, I usually use Tabachnick and Fidell's book along with Cohen et al.'s book as citations for the outliers. I usually also state that I use 2/3 to adequately cover the different properties of regression (leverage, discrepancy, and patterns for Mahalanobis) without being too sensitive by only using one.

    • @niklaskuhme4235
      @niklaskuhme4235 3 года назад +1

      @@StatisticsofDOOM Thank you very much!

  • @torunkaspersen3001
    @torunkaspersen3001 9 лет назад

    Hei Erin
    Tank you so much for this video. It was very helpfull. One question tho. When we tried to make the graph and typed in low, average and high i variable view, it didnt show in dataview when we typed in -1, 0 and 1. It this some kind of setting, or do you think we did something wrong?

  • @nataliewilliams3965
    @nataliewilliams3965 7 лет назад

    Hello Dr. Buchanan. Thank you so much for posting this excellent video! It is very clear and easy to follow. I just wanted to make sure I am understanding my results correctly. My overall model was significant, as was my predictor variable (X) and the moderator (M), however, the interaction itself was not significant. Is it accurate to state the predictor variable (X) and the moderator variable (M) predict Y separately? Also, because the interaction is not significant, it is not appropriate to interpret the simple slopes or the J-N output? Thank you in advance for your guidance!

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      Exactly what you said - x and m predict separately in the main effects, but there is not an interaction. I would not interpret the JN or the simple slope output because they are follow up tests (like ANOVA).

  • @user-iw5xr5ew3f
    @user-iw5xr5ew3f 2 месяца назад

    Hi thank you so for this amazing video, just had a question if in multiple regression (step wise) the IV is not a significant predictor can we still use it in moderation analysis and get significant moderation effects?

    • @StatisticsofDOOM
      @StatisticsofDOOM  2 месяца назад

      Yep! Sometimes it only will show an interaction effect with another variable but not a main effect.

    • @user-iw5xr5ew3f
      @user-iw5xr5ew3f 2 месяца назад

      @@StatisticsofDOOM ok thank you but then how do we take out effect size then as there are is no significant prediction in regression model?

    • @StatisticsofDOOM
      @StatisticsofDOOM  2 месяца назад

      @@user-iw5xr5ew3f I'm not sure what you mean "take out effect size"?

  • @nadinerose-smith8026
    @nadinerose-smith8026 7 лет назад +1

    Hi, thank you for the video, it's great! When calculating a sample size testing more than one moderator, how many predictors do you put in? (I am testing 3 moderators on a linear relationship)

    • @StatisticsofDOOM
      @StatisticsofDOOM  7 лет назад

      I would figure out how many predictors that would be (a lot with all the interactions), and then use that number.