Its still Tukey's HSD. It is just a different way of going about it. Tukey's HSD is really just a t-test where you correct for multiple comparisons and use the MS within from the ANOVA to get a common standard error.

Possibly... The way I handle it is to copy and paste the code it gives in the import window into a script. That way when I want to reupload the data, I can quickly rerun the code instead of having to deal with the menus and windows again and again.

Sometimes it is the simple things.

Without more context, I'm afraid I don't know how to help you. Maybe your looking for the name of your .Rdata file there, and it won't show up there. The R objects that are in the .Rdata file is what you will see.

Glad it was helpful!

Yes, temperature would be a construct. How much heat is in an environment or substance. We measure it not only with different scales (Fahrenheit, calcium, etc.) but also with different tools (mercury thermometer, infrared thermometer, resistance thermometer, etc.). The tool and the scale we use to measure temperature would be the operational definition.

There are a number of different ways you can try. Each have their own pros & cons. For example, you could use concurrent validity where you see if this new operational definition correlates with an already established measure. The ways to try to establish measurement validity is a larger topic that could easily be its own video.

"n" is equal to the number of scores in each condition. There were 5 people in the no phone condition (and the others) and so n=5.

I don't understand you question. Can you be more specific?

You have to look it up in a table. Most statistical textbooks have it in an appendix. I recopied some of the one I use, but I can't show it for copyright reasons.

I would double check that you are matching the right file times to import/opening process.

- {required} Enter name www.psytoolkit.org/doc3.4.4/online-survey-syntax.html Be careful from an IRB perspective, usually participants have the right to skip any question/withdraw.

This is the second of 2 videos (1st one calculates the ANOVA). The dependent variable of driving simulator performance and the manipulation of the independent variable (phone use while driving) was stated at the start of the last video: kzfaq.info/get/bejne/hc-be7R828-roKM.html

@@s.wesleybeckwith3561 Thanks!

What do you mean by subtracting the z-value to 1? Once you calculate the z-score, you shouldn't manipulate it in any way.

They are only crossed to try to show that one line is not "above" the other. The slope of those lines is not intended to be "significantly" different and the crossing can be see as some random noise.

In ggplot, you can add a linear regression line via adding another geom. geom_smooth(method=lm, se=F, color = "#002344", size =1) Setting se to "T" will add the 95% confidence band, and you can edit the color and size arguments to your liking.

They might not have completed it. Check to see how many complete vs. incomplete data files you have. Also, I've found people will try to access them on cell phones, which can do the survey portion but not the behavioral task portion.

All groups have the same n of 5.

@@s.wesleybeckwith3561 What if my groups have different n values?

@@marka2773 Most statistics program apply a correction/approximation technique. R, the stats program I use, uses the Kramer method I believe. To be honest, I've never had to hand calculate it, and don't know as much as I should on this specific topic. Given that Tukeys is a modified t-test, I'd bet the correction looks similar to how an independent samples t-test uses pooled variance. You put MS-error over the n for each group and add them together. I did a little digging and this page by David Howell at UVM says that something along those lines is the case: www.uvm.edu/~statdhtx/StatPages/MultipleComparisons/unequal_ns_and_mult_comp.html Here is an older paper about the subject: link.springer.com/article/10.1007/BF02291420

@@s.wesleybeckwith3561 Thank you both. I was searching the answer to this topic.

The short answer is because you are finding the z-score of a sample mean in a distribution of sample means as opposed to a raw score in a distribution of other raw scores. The longer answer is.... You always calculate a z-score based on the standard deviation from the distribution the score was taken from. So, if you have a raw score taken from a distribution of 15 scores, you need to find the standard deviation and mean of that distribution. Then you can calculate the z-score of that raw score by z= (x-M)/SD. When you are calculating a sample mean's z-score you have to recognize the mean of that sample wasn't actually a score taken from the population. It is a new value created by calculating the mean. So the mean didn't really come from the population but actually from a different distribution. The distribution it came from is the distribution of sample means. This is a new distribution comprised of every possible sample mean from a population for a given sample size (yes, it is a mouthful; basically-but not exactly-all the possible sample means). From that point on, you are really just calculating a z-score. You take your sample mean and subtract the mean of all the possible sample means (i.e., the mean of the dist. of sample means). You take that difference and divided it by the standard deviation of all the possible sample means (i.e., the SD of the dist of sample means). The Standard deviation of the distribution of sample means IS the standard error of the mean. So in calculating the z-score of a sample mean, you are using the standard deviation in standard error because standard error is a specific standard deviation that is so important it gets its own name. If you are asking this question, you probably need a refresher or an introduction to the distribution of sample means. I have a video on it. It is older and needs to be rerecorded, but I've linked it anyway. I hope this helps: kzfaq.info/get/bejne/l9WZidmHzNXFoZs.html