I wanted to respond, "Just increase your degrees of freedom! That's what I do."From: Ducerveau Demerde
To: Angry Professor
Subject: Advice
Dear Dr. Professor,
I read your paper ["Modern Statistical Techniques for the Analysis of Measurements: Why ANOVA Sucks and How You Can Avoid It"] recently and it was very helpful as I was trying to understand my data. I am currently confused by something and I really need your help.
My area of research is [Frybread Studies] and my dissertation examines [relative utility of maple syrups absorbed in pancakes, crepes and waffles]. I have run a series of experiments in which I collected measurements that I wish to analyze using ANOVA. According to my hypothesis, there should be an interaction between type of syrup and type of bread. Unfortunately, pancakes, waffles and crepes absorb syrup differently, leading to a great deal of variance in my measurements, and the interaction is not significant.
I have read all of your papers, the one I cited above, as well as ["ANOVA: The Rape of a Social Science"], ["P-values Are Meaningless"] and ["You Can Do Better Than ANOVA"], and based on what you say in them, I think I can process my data differently so that the interaction (via ANOVA) is significant: I can use the inverse of the measurements to reduce the variance. I have used SAS, SPSS, and SYSTAT, but even using the inverse, the interaction is still not significant. However, I found another program online that seems promising (it is attached) and I have several questions that I hope you can answer.
<snip an additional page of questions about preparing the data for an ANOVA>
Thank you for taking the time to read this, and I hope you can help me because if this interaction is not significant my dissertation will be a disaster.
Sincerely,
Ducerveau
<snip attachment of several megabytes of executable incompatible with my operating system and an Excel spreadsheet containing his raw data>
But no. Despite my aggravation, I responded very politely. I can understand the desperation of a graduate student thinking his dissertation experiments are for shit, but I can't understand how anyone who has read my work could even dream of sending me this letter.
10 comments:
Get well soon.
What's the inverse of an ANOVA? Are you serious about that point, or was that all folded in with the anonymization of the post?
Hot lemonade and honey. That -- and "42", are the answers to any questions in the Universe. Plus, the former will make you feel much better (the vitamin C is good for you, the warmth soothes the throat, the honey ("gift of the gods") provides some energy. The latter? When that dolt achieves that age, much will be clearer.
Ah, academic Christmas cards. I hope you're feeling better soon.
And if anyone has any advice on dealing wtih an unrepentant racist who likes to come in and talk to me about his father's Klan meetings, I'm all ears. He told me today that he'll see me in the spring.
I think brandy needs to be added to that hot lemonade and honey, and then you are good to go.
"ANOVA: The Rape of a Social Science" .... love it!
I hope you feel better. As you know already, I'm a big supporter of equal parts of honey, lemon juice, and whiskey -- topped off with some hot tea.
It sucks to be feeling puny at this time of year. There is always the external family stress to deal with and the pressure to look like you are feeling good/having a good time in front of the Angry Kid. Try to buid in the new tradition of Holiday Nap Time and Central Nightime Savings Time -- meaning bedtime at 9:00 p.m.
Dog bless and get well.
Aidan
Ha! Should have told the student to change their study to embrace the heteroscedasticity in the data-- maybe some hypotheses about the differences in rate of absorption of syrup by pancakes, crepes, and waffles. It will totally result in a world-changing dissertation, like all of ours were.
Kind of annoying of me I realize, but after seeking sanity in other prof's blogs, I've started my own. You might enjoy it Gentleman C.
http://acadamnit.blogspot.com/
Better than increasing the degrees of freedom, report a p of 0.42 as a 'trend' and discuss it in the rest of the paper. I great idea I got from manuscripts I review.
Sometimes I wish this wasn't pseudonymous because I think I would like to read "P values are meaningless" and "ANOVA: Rape of a Social Science."
Also I think that anyone who thinks that simply changing the stats program you are using might result in an increased effect size does not deserve to pass their dis defense.
Post a Comment