One-way ANCOVA on SPSS with assumption checks and write up (APA style) 

Thank you!

Thanks Jeff, glad it helped!

Glad it helped!

Thanks, glad the videos are helping!

Hi, the test is usually considered quite robust to violations of normality, so these wouldn’t rule the test out, but violations could be acknowledged in your report. Re. the assumption that there’s a linear relationship between the DV and the covariate, if this is violated it can reduce the power of the test. However, again, this wouldn’t necessarily rule out running the test. In this case, you could consider removing the covariate, which would mean running an alternative test (e.g., a t-test) if there was only one covariate. If the assumption of homogeneity of regression slopes is violated, the test results may be misleading, so you may need to consider an alternative test. Regarding homogeneity of variance, you can adjust your alpha (e.g., from 0.05 to 0.01) to help account for violations. Hope that helps!

Thanks Saumya, glad it helped!

Thanks Jennifer, glad it was helpful!

Thanks, glad it helped!

Thanks, glad it helped!

Glad it helped!

Hi, thanks for your question. I haven't seen any resources that they that the covariate has to be checked for these assumptions too.

Glad I could help!

Hi Kim, Thanks for your question. Generally speaking, researchers tend to use the Kolmogorov-Smirnov test with larger samples (n > 50) and use the Shapiro-Wilk test with smaller samples (n < 50). Here's a reference: www.ncbi.nlm.nih.gov/pmc/articles/PMC6350423/#:~:text=The%20Shapiro%E2%80%93Wilk%20test%20is,taken%20from%20normal%20distributed%20population. Hope that helps! David.

Hi, yes, ANCOVAs are often used to do just that.

@@DavidRobinsonPhD thank you very much Dr

Hi, Thanks for your question. It's fine to have multiple covariates. It's more common for the covariate to be continuous, though many argue that it's fine to have categorical covariates, and that the assumption of linearity doesn't apply in this case: www.researchgate.net/post/Check-linearity-between-the-dependent-and-dummy-coded-variables Hope that helps! David.

Hi Lowie, there's a downloadable dataset here that you could use: help.xlstat.com/s/article/ancova-analysis-in-excel-tutorial?language=en_US

That's just to talk to people outside the building 😅

@David Robinson, PhD that makes more sense🤣🤣

Hi Aina, Thanks for your question. Could you let me know what your IV(s), DV(s), and covariate(s) are? Best, David.

@@DavidRobinsonPhD Hi David, thanks for very informative videos and really making data analysis relatively easier.When reporting the results, you state that the marginal means were similar (I can appreciate that visually since its 5.82 vs 5.83). Is this based on the p value suggesting there was not a significant effect of condition on cognitive scores? How can make a such a judgement statistically? Do I use contrasts or post hoc tests?

@@mwanganamubita9617 Hi Mwangana, Thanks for your question. Yes, it's based on the non-significant effect of condition on cognitive scores. Also, the partial eta squared value (.003) suggests a very weak effect. If you wanted to avoid using "similar", you could say something like, "...there was not a significant difference in cognitive test scores between the control [insert descriptive stats] and intervention [insert descriptive stats] conditions, [insert inferential stats]." Re. post-hoc tests, they're usually only used when your IV has three or more levels. Hope that helps! David.

@@DavidRobinsonPhD Many thanks for the clarification

@@DavidRobinsonPhD, Do you have a video on repeated measures Ancova or can you help me with how to report results of repeated measures Ancova with dependent variable at two levels (measured twice)? Warm regards