Dear colleagues,
As the traditional t-test statistics (T) is strictly related to point biserial coefficient of correlation (RPB), and because RPB is prone to give estimates that are radically deflated, consequently, T must be deflated. I studied a bit the magnitude of the deflation in real-life settings and introduced a new test statistic, deflation-corrected t-test statistic (DCT). In DCT, the deflation-prone coefficient RPB is replaced by attenuation-corrected RPB (RAC). I cordially invite you to read the paper and give ideas how to enhance the text. Any kind of ideas would be appreciated.
Regards,
Jari
mix discrete and continous data is something called logistic distributions. i remember the veteran teacher on my uni gave me a pdf about this kind of situations. Its not imposible to handle this kind of spaces are precise and hard. You need to read about discrete spaces, categories or cualitative distributions and convolution of spaces. Its more about the equivalent tool for discrete variables and use the same theory to deduce the formulas, good after.
To follow up on Dr Caceres reply,you might look at the references in the attached paper or even just Google logistic regression Best wishes David Booth
Hej Jari,
Eftersom Wilcoxon-matched pairs signed-ranks test anges med T (Siegel, 1956) så stör det mig att Student t-test skrivs med stort T. Detta är endast en liten parentes.
MVH
Béatrice
.
Hej Béatrice! Ursäcka min svag Svenska... You are correct. T is also used, among others, in Hotelling's T-distribution, as well as sometimes when tau is used (e.g., with "true" score in measurement modelling or Goodman - Kruskal tau). Sometimes T is used also with t-test but rarely except when it starts the sentence. I need to rethink this. Let see how the reviewers react.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size