I had no idea you were such a prolific RU-vid contributor. I'm glad I stumbled upon your creations. Lots of viewing for me now during the pandemic. Keep up the good work.
Hello, I am testing a seasonal variation of different metals concentrations in the fish liver. I am testing 9 groups (metals) in total for four seasons. For the first and for the second season I have results for 5 fish (n=5), and for the third and fourth season, I have results for 3 fish only (n=3). All four groups (seasons) have normally distributed data. I have conducted Leven's test, and 5 of 9 metals have unequal variance, while the other 4 metals have equal variance. Should I apply diferent post-hoc tests, i.e. for the 5 metals with unequal variance and unequal sample size to choose one post-hos test, and for the rest 4 metals which have equal variances and unequal sample sizes to choose the other post-hoc? Which post-hos tests would you recommend in my case? I can send you analyses of these tests and data if something is not clear enough. Thank you for your response in advance.
Hello, firstly thanks for your video;very helpful. I was wondering if the same could be done for a Scatter graph or its not necessary to do all the steps? Thanks
Hi, Swati - you cannot include control variables in a t-test; instead, you would conduct a repeated measures ANCOVA (even though your IV has just 2 levels)
Thank you, Doug. What if I'm comparing mean scores of two people in a dyad or pair situation. For example, lead and assistant teachers' wellbeing. I am trying to control for both teachers' "experience" teaching preschool. In this scenario, can I still use ANCOVA? In this situation my IV is Teacher type. Levels are lead or assistant teacher.
That case might be trickier because you have two separate levels of experience for each teacher rather than a single control variable for the pair (unless you averaged their scores on experience). You might need to consider multilevel modeling/dyadic data analysis which is beyond my scope of knowledge. But perhaps this would be a start: davidakenny.net/dyad.htm
Nice video, but it doesn't have to happen in this way. SPSS is now at version 24, costing hundreds of $ per year. Why it can't get the simple stuff right? why still lacks critical functions?
I've conducted a planned contrast and all of my values for the t column are negative, what does this mean and do I include this - sign whilst reporting my results?
The sign indicates the direction of the difference - negative just means that the (average) mean for the second group is higher than the (average) mean for the first group. Reporting the sign is optional - just make sure it's clear to the reader what the direction of the mean difference is.
Hi Doug, I was wondering if you could help me. I am comparing 3 groups. It is a between subjects design, where I am comparing group 1: postal reminder for dental appointment group 2:text message reminder for dental appointment and a control group: no reminder I believe I use a one way ANOVA to find out whether the results are significant, however, I do understand if I would use a post hoc test and was wondering if you could help me? thanks! Michaela
+Mixed Peas Hi, Michaela - You are right that a One-Way ANOVA is appropriate in your case. Post hoc tests are possible (if you get a significant F from your ANOVA), but if you have a specific hypothesis to test, planned contrasts would be a more powerful approach (see my other video on them). Best of luck!
+Doug Maynard many thanks for your help! Due to my experimental research using categorical data I have now chosen a chi-square test of analysis. I am comparing (IV) 2 intervention groups to a control group (so 3 groups) and my outcome measure/ DV is attendance rates. I was wondering if you have any clips describing how to deal with chi-square data? E.g.: what is the difference between the goodness of fit and contingency chi-square statistical texts as I am not sure which one to go for. Many thanks, Michaela
+kevin rollins Sure, that would be fine. It would be equivalent to the 1, -.5, -.5 that I did here, meaning that you should get the same t, df, and p results. So long as the contrast is balanced, the actual values are up to you.
+Driton Vela It's a good question. There are no hard and fast guidelines that I know of, unfortunately. I think you would probably be safe with either the Gabriel or Hochberg's in this case.