Hiya, I have a set of data with one IV (3 levels) and 1 DV which fail a Bartlett's test for homogeneity of variance and the data cannot be transformed, so I perform a Welch's ANOVA, but cannot find how to perform Games-Howell post-hoc tests in R. I have a 2x4 factorial between-subjects design and would like to calculate post- hoc tests using the Games-Howell procedure in SPSS. I have been doing some research and have found that the games-howell post hoc test can deal with this type of data but can not find any code/ algorithm for it in R. Does any one know of the code or a different post hoc test that will do the same thing but is supported in R?
Games Howell Test Spss 1
The Theory of Poker Applied to No-Limit now availableFor those of you here in Las Vegas, The Theory of Poker Applied to No-Limit by David Sklansky is now available at Gambler’s General Store/ GAMBLER'S BOOK CLUB in downtown Las Vegas. Their address is 727 S Main St, Las Vegas, NV 89101 and their phone number is (702) 382-9903.We also have this title available in several special directly from Two Plus Two Publishing.For more info or to ask questions check out this thread in the books and publications forum:.Probability Discussions of probability theory. Hello again, 2+2 SMPers. As 2+2 always provide excellent answers to my inquiries I'll give it a go again.This thread will hopefully point me in the right direction of understanding, or even produce an answer, to my question, as well as discuss how tragically bad understanding of statiscs are.Those who cba.
To give any advice skip to B).A)This is the case:We wish to look at how different environmental poisons affect fish (in this study ossification of bones - bone growing) (ossification of bones are just a part of the study, we also have a set of genes we wish to investigate)We have a control, a low exposed group, and a high exposed group.Data is collected on day 2,5,7 and 11 (# of bones)In each group there are 10 individuals (for each day, means 40 in total)This produces a dataset which is to be analyzed. Is there a significant difference in number of bones between the groups?We are now told to perform a one-way ANOVA, and a post-hoc test (Tukey).My understanding is that the one-way ANOVA will tell us if we can reject the null hypothesis or not, while the post-hoc test (Tukey) tells us which results are responsible for it.SPSS flags the differences that are significant.Data produces by the Tukey-testLooking at data for the Pectoral bone on day 7. First let me state that I know very little about this, but read it with interest. Not a statistics expert, not even a statistics amateur, although I have done a little bit of research and used SPSS before. Now that thats said, questions.0.
Hypothesis:H0: Poison A does not affect ossification.H1: Poison A does increase/decrease ossification. Yes, yes, yes, Looks to be the homogeneous ballpark by eyeball test. Unless you perform a test of homogeneity you don't know that the data are significantly non-homogenous. You want to use Tukey if you can. Games-Howell test loses some power, but should be used if data is proved not to be homogeneous. Error bar is 1 sd and has some use. Not sure what your saying about calculus/statistics.
Statistics is an entire field on its own and can entail as many courses (20) by itself.Maybe should be in probability forum. First, what does the overall anova look like? What response variable did you use? It looks like you did several individual anovas for day a, b, c, and d from the same individuals? This is not kosher as they are not independent. A rmanova would be appropriate. As for post hoc tests, I tend to use an SNK for the most part, since it's a reasonable balance between too liberal (LSD) and too conservative (bonferroni, scheffe).Error bars on a graph should never be used to judge significance.
Typically, error bars of +/- 1 se is the norm, not sd. However, graphs do help to explain results. In your case, it seems that control = low. Basically every statistics test and professor gets this wrong.
What you want to do is an ANOVA with a linear contrast.not an overall omnibus test with a bunch of post hocs. You want one analysis that directly tests your theory that increased exposure leads to increased bone loss.I am not 100% sure how to do this in SPSS off the top of my head, but you want to click on a 'contrasts' tab at some point and set the contrast weights to -1, 0, +1 for the three levels of exposure respectively. It should give you an F-contrast and a p-value that tests your theory directly, which is way better than doing all possible comparisons as post-hocs do.As for the problem of unequal variances in the group, the resolution for this issue is not entirely clear. There are adjustments for the p-values and f-values that SPSS can provide based on unequal variances (Hyundt-Felman, Geissler-Gaussen or something like those), but even those leave one with further issues iirc.Type in 'contrasts and effect sizes in behavioral research' into google.
When I do it, the 3rd item down gets you to this:which should be helpful. Hello again, 2+2 SMPers. As 2+2 always provide excellent answers to my inquiries I'll give it a go again.This thread will hopefully point me in the right direction of understanding, or even produce an answer, to my question, as well as discuss how tragically bad understanding of statiscs are.Those who cba. To give any advice skip to B).A)This is the case:We wish to look at how different environmental poisons affect fish (in this study ossification of bones - bone growing) (ossification of bones are just a part of the study, we also have a set of genes we wish to investigate)We have a control, a low exposed group, and a high exposed group.Data is collected on day 2,5,7 and 11 (# of bones)In each group there are 10 individuals (for each day, means 40 in total)This produces a dataset which is to be analyzed. Is there a significant difference in number of bones between the groups?We are now told to perform a one-way ANOVA, and a post-hoc test (Tukey).My understanding is that the one-way ANOVA will tell us if we can reject the null hypothesis or not, while the post-hoc test (Tukey) tells us which results are responsible for it.SPSS flags the differences that are significant.Data produces by the Tukey-testLooking at data for the Pectoral bone on day 7. Thanks a lot for the great responses!Tao1, you adress very interesting questions. Bone ossification is just a part of the study (a rather small one).
We are more interested in a set of genes, which in turn is a lot more complicated to process.Basis for the high/low doses are earlier studies. (It's polychlorinated biphenyl congener 77 - PCB by the way) This is mechanical toxicology, and we expect that there will be a significant difference for the high dose.Pokerlogist, I understand I didn't make myself clear. What I'm trying to ask is; Why aren't we learning statistics instead of calculus?
I find more real world applications for statistics than what we learn in calculus (I've done up to multivariable, and complex analysis)Zoltan, thank you very much for your advice (and link)Sherman, thanks a lot! Will do some reading about thisVBAces, I will prolly take you up on your offerYou've all been really helpful!
Games-HowellLab ExerciseMultipleComparisons-cont. 1/25/2001Solomon, Secker-Walker, Skelly, and Flynn (1996) Journal ofBehavioral Medicine studied smoking behavior in pregnant women. They looked at thewomens determination to quit smoking while pregnant. They interviewed 349 women attheir first pre-natal visit, all of who were smokers when they became pregnant, andclassified them into four groups. PC Precontemplation Smokes and has no plan to quit smoking.
C Contemplation Smokes but is thinking of quitting. P Preparation Smokes, but has made some effort at quitting. A Action Has already quitThey wanted to look at the subsequent smoking behavior of thesesubjects over the course of their pregnancy, but one important consideration is how muchthese women smoked when they became pregnant. If the groups differ on that variable, thatmight affect the interpretation of the results.
(This is our problem for today.) The datacan be found inThe answers to this exercise may be found at.The means and standard deviations of these four groups, in terms ofcigarettes/day when they became pregnant, follow. Dev.13.35.212.28.8n j693715390Notice that this is not really a simple problem. Your sample sizes aregrossly unequal, and you have problems with heterogeneity of variance. Not to worry, Itell you how to deal with this in the text using the Games-Howell approach. It is amodification of the S-N-K, though can be applied in the context of most pairwisecomparisons by making suitable changes in the critical value of q r.Notice that this is a real data set, and this is the kind of problemthat each of you can expect to face in the future. This isnt some trumped up examplethat doesnt apply to anything important. Using the Games-Howell procedurewithin SPSS, what canyou conclude from these data?
Tell me how this test differs from a standardS-N-K or Tukey test, as far as the arithmetic is concerned.Last revised: 04/10/03.