Search results

1. Linear Regression with Negative Response Variable

it mihgt be productive to, instead of taking linear regressino of difference = ..., to the variable being subtracted to the rhs, ie Y1 - Y2 = mx + b ---> Y1 = Y2 + mx + b, ANCOVA sort of style. or just scatter plto Y1 x Y2, would be revealing.
2. Which test should i use ?

with n=3, you should just plot, there's no real point in testing. You can seen how it is. You have a small n big p problem it usually called. In such cases 'dimension reduction' can be desirable, ie to compute a sum or averge over the 4 independant variables to get one 'very imporatnt...
3. Mixed ANOVA Unequal Gender Ratio

yeah i think your worried about something you shouldn't worry about. It depends what hypotheses you are testing, but in pretty much any practical situation, its a non-issue.
4. Which test should i use ?

Probably just run a t-tes on each of the 4 independant vars, one at a time. If you want to test all 4 at once, ie to test the hypotheis of any difference across all 4 vars, then hotelling t^2 i think is the test. Hotelling's T-squared distribution - Wikipedia Scatter plot matrix is also a must...
5. basic question: surely that his graph is visually misrepresenting the data?

i think it just depends how 'risk score' was derived defined. the confidence limits look symmetric on the given scale, that may provide some clue.
6. Mixed ANOVA Unequal Gender Ratio

I think if gender is independent then it is not a real issue, except possible there is some reduction in the power of some tests. What do you mean by 'makeup for this discrepancy'?
7. joint probability distribution three variables

fine what is mutually exclusive here, you mean they are binary and Prob( x=1 and Y=1 ) = 0, or what?
8. An estimator for a Cumulative distribution function

for 2. you will have to compute the mean and variance of Y, by taking appropriate integrals, then set them equal to the sample moments Y_bar and variance of Y. You'll probably have to differentiate with respect to y to get the pdf. see how that goes....
9. Which statistical test?

does have a sort of 'ordinal' character to it, ie increasing severity?
10. Dream's Speedruns

The important thing is that the interested parties agree a-priori (before hand) how the stats are going to be counted, this is extreeeeeeeemly important when in any sort of adversarial context. Its arguably more important than 'doing it right'. What outcomes will be considered 'cheating'...
11. Dream's Speedruns

just a general comment on these issues of 'cheater' catching. Need to keep in mind that the important part is the 'positive predictive value' ie the chance that they are cheating given the observed outcome, rather than the observed outcome given the evidence, which is what is often tabulated...
12. What is a general linear model?

I feel like there should be some sort of convention for naming stats procedures, like IUPAC does for chemicals. I think multiple regression is really just a subset of general linear models, but software often gives different procedures for the two, which is where this language is coming from.
13. Bernoulli's Trials / HELP!

its the bracket looking thing in the formula: Binomial coefficient - Wikipedia
14. How to transform trip generation data into normality using SPSS

Hypotheses non fingo
15. How to transform trip generation data into normality using SPSS

classic answer would be 'box-cox' transform? SPSS likely has a button for this...
16. Help analysing variance among groups

usually 'folded f test' to compare two sample variances. 1.3.5.9. F-Test for Equality of Two Variances (nist.gov)
17. MANOVA, multivariate regression or paired t-tests? help please :)

I had a similar issue recently dealing with this sort of issue of estimating correlation in a repeated measures setting. I was just going to mention that I found the multilevel modeling approach mentioned above computationally unweildy, at least in SAS. I also saw some references thatsuggested...
18. ABOUT Z TESTING, HYPOTHESIS TESTING AND α

Any one of those is going to make a more readable book than t-tests. World needs another t-test book=not really.

chi square
20. Old Maid

very unlikely id say. Basically you got '37 choose 19' possible deals, assuming you get the more cards. For each possible pair you can get one of the possible pairs, so there are 2^18 deals that give no pairs. Something like that.... So ur looking at 2^18 / choose(37, 18 )... or like, 0 or 1...