Normalization procedures
Hello everybody, Has anybody heard of the Bell-Doksum test? If so could you please give me a reference or a short descriptiion of it. Some one mentioned it to me and it's driving me crazy not to be able to find any info either on the net or in my references. Best Niko Tiliopoulos = Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at http://jse.stat.ncsu.edu/ =
Re: which test to use
Dear Kathy, You slightly confuse me with all that detail, but if what I get is right, and that is that you have two continuous variables (one IV one DV), then why don't you use a simple regression analysis? Is there something I overlooked or does this appear to solve your query? Best Niko Tiliopoulos = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: When one should stop sampling?
Dear all, In light of the very interesting and highly appreciated response I received in my mailbox, allow me to attempt to be more clear. First I should say that I am not aware of the deep details of the study (it is indeed someone else's and I am not trying to cover up my errors). Ss are put into a half an hour or so deep relaxation state in a soundproof environment and are asked to speak loud their thoughts. After that they are presented with four photographs and asked to rate them for their similarity to what they were thinking during the relaxation stage. The photos are pre-selected by the computer and the experimenter is blind to their content. The computer has also pre-selected and registered the winning photo (again the researcher knows only that this photo is one of the presented four, but not which one). After these two sessions the subject fills in a few questionnaires and the testing ends. What she is trying to see is whether the Ss have an above chance hit on the winning photo (though this is probably very simplistically stated). Apparently there is a foundation behind this rationale, but I am not familiar with it, and frankly I do not think it is relevant to the issue raised. Also there may be more conditions in the study that I am not aware of, but once again I do not think they matter here. Finally, I know that there was a pilot prior to the study. The intended sample of 200 was stated in an approved grand proposal and in the proposal to the ethics committee. And as I mention in my previous posting, potential Ss are very precious since they have to fulfil certain requirements. The potential flaw has to do with a condition during the relaxation session. Drumbeats are used as a background noise because some evidence suggests that this harmonises with ones heartbeat. However, Ss tend to complain that it distracts their attention or that instead of relaxing them, it produces the opposite effect (I trust the volume levels have been checked). Should this be the case, then this may be the (a) reason why she is not getting the expected effect. Therefore she is considering changing this to simple and well-tested white noise (don't ask me why, but there has to be some background sound). Initially I had suggested to split-half her intended sample and randomly assign half of the Ss to the drumbeat condition and the other half to the white noise one. However, when I was told that 40 people had already been sampled and that the results were really (and perhaps surprisingly) discouraging, the issue shifted to the one I presented here in my initial post. I would like to make two additional comments/clarifications: In response to a mail from Dennis who wrote: if she has run 40 Ss with no results ... that certainly will NOT get published , I ensure you that from what I know of that area (precognition, PK, etc.), no results are as good (and indeed much more common) as some results - though I will agree that one should perhaps be more concerned with getting positive and replicable results. However, in this particular instance (only) I find it hard to subscribe to the view that perhaps it is still too early for an effect to show up and that perhaps she should indeed meet the 200 target and check again. The thought of this being an unnecessary waste of valuable resources really scares me. Finally replying to Donald's mail who said: we keep being reminded that the null hypothesis is never actually true, which implies that the ES is not exactly zero, which implies that with a sufficient sample size (maybe ten million or so?) the power curve would indeed level out -- near power = 1.0.) If one wanted to invoke a statistical argument (in the face of whatever logical argument and/or evidence exists of a design flaw and/or of an ES an order of magnitude smaller than one had reason to expect in the beginning), it might be more persuasive to show that an upper bound on ES (say, the top of a 95% confidence interval) would imply no practical value whatever for so small an ES. (Presumably the presence of an interestingly large ES would have implied some change, or recommendation for change, in practice somewhere.) I do agree with all of this, and in fact, as I say to my students, if I could have a million participants I would probably be able to detect a significant association between, say, schizophrenia and banana consumption. However, I was hoping in this particular instance that the power curve would show signs of levelling already with a sample size of, say, 100 Ss (perhaps too optimistic, I know). Once again I do appreciate your input, and I hope you forgive my unintended verbosity. Best Niko Tiliopoulos Department of Psychology The University of Edinburgh 7 George Square Edinburgh, EH8 9JZ = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available
When one should stop sampling?
Dear all, I am acting as the stats advisor for my unit in the psychology department of the University of Edinburgh, UK. Last week a colleague of mine presented me with the following issue, and I am not quite sure how to respond: She is running a psychological experiment, in which she a priori specified her sample size as 200 people. She has already sampled 40 participants and a preliminary effect size (ES) analysis suggests an almost zero effect. Based on previous research, she was expecting a detectable effect even with 40 subjects - though I suspect she was not expecting enough power to get a significant result at that stage. In addition, it appears that the reason the ES she gets is nowhere close to the expected figure may be because of a design flaw. So she asked me whether it is justified to go up to, say, 100 participants, check again her ES and if it's still near zero, stop sampling, or whether she had to sample all 200 people because she had says so in her protocol? I do think it would be foolish to keep sampling when one has grounds to believe that there is no effect or that there is a flaw in the study. I believe that if the plot of subjects versus power, suggested that the power curve levelled after a given sample size, that would be enough justification to stop sampling (needless to say that participants that satisfy her protocol are precious and hard to find). Her query though sounds to me more like a methodological (if not ethical) one, rather than a true statistical problem, and thus this bottom-up justification may not suffice. However, I have failed to think of a top-down, theoretical approach to this problem, and since she will try to publish the results of this study regardless, she would like references that back up my position (or in that sense falsify it). I would really appreciate your comments on this. Best Niko Tiliopoulos Department of Psychology The University of Edinburgh 7 George Square Edinburgh, EH8 9JZ = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: MDS, the radex, and indices of multidimensionality agreement
Thanks a lot John, Indeed Gutman has worked on the radex, so I get something out of his papers. I also found your suggestions on the agreement very helpful (haven't tried them yet). Once again cheers Niko Tiliopoulos Psychology Research The University of Edinburgh Uk = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
MDS, the radex, and indices of multidimensionality agreement
Dear all, I was not sure in which of the forums the following two queries should be posted, so I have put them in all of them - I do apologise if they are not appropriate for the current list. Q1. I have run a multidimensional scaling analysis (MDS) and the 2D-map suggests that the variables are arranged in a circular-like fashion. I have found a paper that presents a 2D-map showing a similar arrangement. In that paper the author has drawn on the 2D-map a number radii and two concentric circles that divide the variables in theoretically meaningful regions (e.g. the area between two radii may represent a specific psychological trait). This whole structure is called (in that paper) the radex. However, it is not mentioned how these lines were drawn and my attempts to find out were so far fruitless. Does anyone know the formulae I can use to draw these lines, or whether there is a software package that can do it for me? Q2. I have also run a factor analysis on the same dataset, and I would like to compare the level of agreement between the FA factors and the MDS dimensions. Has anyone come across the coefficient of congruence and the Borg-Leutner alienation coefficient? If so, I would be obliged if you tell me where I can find their formulae. Thank you in advance Niko Tiliopoulos Psychology Research The University of Edinburgh UK = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Alscal vs. NCSS
Dear all, I have two questions regarding MDS: 1. I have run an NMDS through Alscal (SPSS) and NCSS, and the representations of the variables on a 2-dimensional map look completely different. As far as I can tell, I am using the same procedure in both algorhythms, so I cannot understand why I get different results, and which one I should prefer as more accurate. 2. Does anyone know which of the following two stress indices should be used with data from psychometric instruments (e.g. personality questionnaire): Kruskal's or Guttman-Lingoes? Thank you Niko Tiliopoulos = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =