New answers tagged

1

Estimates of $\mu$ and $\sigma$ will necessarily be rough with observations spaced so far apart. You might guess that $\sigma = (144-120)/6 = 4.$ If you saw about half of the $n=15$ were degraded by hour $132$ then you might suppose the population mean (also median) must be about $132.$ However, with only five of fifteen degraded (about $1/3),$ Then you ...


1

Use classification when the number of categories are limited and nothing in between makes sense. For example a class is either a dog or a cat nothing in between. But when it comes to something like ratings, 3 is as likely acceptable as 3.5 ((so is 3.56, etc) so you are not bound to only one value among others. it can be in between as well. Apart from this, ...


1

It is OK to stratify a continuous predictor to deal with violation of proportional hazards (PH). Harrell says on page 501 of the second edition of Regression Modeling Strategies: When a factor violates the PH assumption and a test of association is not needed, the factor can be adjusted for through stratification ... For continuous predictors, one may want ...


2

You computed the CDF by using the proper integral of the PDF $$\int 2x^{-2} dx = \frac{-2}{x} + C $$ But what you forgot is to use the correct integration constant (or use a definite integral). Your CDF is not $$F(x) = \frac{-2}{x}$$ But instead $$F(x) = \begin{cases} 0 &\quad \text{if} \quad x \leq 2 \\ \int_2^x 2u^{-2} du = 2 - \frac{2}{x} &\quad ...


2

Don't bin your continuous data. Feed them into your algorithm as-is; potentially transform them using (e.g.) restricted cubic splines (see, e.g., Frank Harrell's Regression Modeling Strategies) to capture any nonlinearity. In particular, don't go hunting for significance by "adjusting" bins. Your $p$ values will be biased low. This is no different ...


2

No. In general, regression modeling makes no assumptions about the distribution of the predictor variables (except, for some applications, that the predictor variables are observed without error). The choice of the response distribution (e.g. Gaussian vs Poisson vs Gamma, zero-inflated or not, ...) is all about the conditional distribution of the response ...


1

The best way is not to filter "outliers" at all What we call "outliers" in statistical analysis are points that are distant from the majority of the other points in a distribution. Diagnosis of an "outlier" is done by making a comparison to an assumed distributional form, and statistical tests for outliers compare the position ...


Top 50 recent answers are included