Analysis of Covariance (ANCOVA) - SPSS (part 2)

[@KineKeys]

Have I told you lately that I LOVE YOU! Thanks for these wonderful tutorials. Very clear and easy to understand. What a great refresher.

[@Frankligia]

Statistics in plain English! Thank you for explaining so clearly rather than using your video as a showcase for your statistical knowledge as some other stats videos do!

[@peterbolin2720]

Can we please have closed captioning on the video enabled? The automatic transcript option is not showing up for some reason....

[@capdepus]

Thanks for so clear explanations!

[@omer197435]

Can we get the real data so that we can do the analysis after watching the videos (parts1-4)?

[@AJP0987654321]

you're great

[@how2stats]

Statistics is the language of love.

[@nikmoanikamo4201]

I mean not sure if you have interpreted eta partial squared correctly in this video. Otherwise, the video is very helpful. Thanks

[@nikmoanikamo4201]

sorry, it is ok, because in a case of one-way ANOVA partial or full eta squared would be the same! :-)

[@pinoyako1927]

Thank you so much for this tutorial sir. Where can we access the dataset for our practice. thank you.

[@how2stats]

I only make available the data files for my textbook available here: www.how2statsbook.com

[@meditation444]

Why did you assess the homogeneity of variance? I thought you could only use it when comparing three or more groups?

[@how2stats]

Definitely not the case. Homogeneity of variance can and should be assessed even with just two groups. In fact, SPSS tests homogeneity of variance automatically in the independent t-test case.

[@holypicklesmofo]

@@how2stats What would you do if this assumption is not met and the Levene's test p < .000? I'm trying to assess some differences between smokers and non-smokers with big N differences (65 smokers and 586 non-smokers).

[@hannahalveshagy4141]

When would I not want to create a composite score? I tried to use the dimension reduction technique to create a composite score for three demographic variables. Is that wrong?

[@how2stats]

It is probably wrong. You should consider creating a composite score when the variables are inter-correlated positively; ideally, they are inter-correlated positively sufficiently to yield a coefficient (Cronbach's) alpha of .70 or greater.

[@hannahalveshagy4141]

@@how2stats Thank you so much for your help!

[@deniseperri6707]

If I want to know if the respondent's sexuality has an effect on the decline of brand attitude between a pre-test and a post-test, would ANCOVA be a good option? I did a repeated measures ANOVA to see if my between-subject factor had an effect on the difference between my within-subject factors but now I still need to see if variables like degree and sexuality can be the cause of this. Not sure if this is clear but I thought I'd give it a go.

[@how2stats]

Sounds like a mediation analysis. I don't think I have a video on that, yet, but it will be in my upcoming textbook.

[@deniseperri6707]

Thank you for your answer. Is there any video on this analysis you can suggest?

[@deniseperri6707]

Also, I now put 'degree', 'sexual orientation' and 'attraction to model' in the 'covariate' box to see if these variables covariate with the changing brand attitude. Do you think this is a plausible solution for my question? I can now see in the output that for example 'attraction to model' is a factor that has an influence on the relation between the dependent and independent variable.

[@icarocosta4302]

my teacher, a little help here. and if my levene test is p < 0,05 ?

[@how2stats]

Are your sample sizes equal? If not, which sample has the biggest variance?

[@WalyB01]

what lets do and PCA? This is getting advance.