nope
  • Member for 3 years, 2 months
  • Last seen this week
Why doesn’t a $F$-statistic of 1 result in a $p$-value of about 0.5?
Accepted answer
5 votes

I think this chain of reasoning contains a bit of confusion between mean and median. A p-value of 0.5 should be expected at the theoretical median (i.e. $F^{-1}(0.5)$), not the expected value. From ...

View answer
Highly correlated engineered features any helpful?
4 votes

Given that car model is a categorical variable, replacing it with some numerical variables such as horsepower or mpg can have advantages depending on the scope of your prediction model. If you only ...

View answer
Why sd of a normal distribution changes by only changing its mean (R code provided)
Accepted answer
3 votes

$N$ containts the realisations of 10000 random variables, each with a different mean. Therefore, the observed deviations from the average are higher. Same goes for $M,K$. If you calculate sd(N-U), ...

View answer
why does lasso select at most n predictors?
Accepted answer
3 votes

Consider a linear model $Y = X\beta + \varepsilon$ with $p$ variables and $n$ observations, $p>n$. Assuming the variables are not linear dependent, i.e. the matrix $X$ has rank $n$, $Y$ can be ...

View answer
How to interprete likelihood equal to 1
Accepted answer
3 votes

Likelihoods express the probability, that the observed event (two heads up) occurs given some parameters ($p = 1$). If you have a coin that always lands on head, two heads in a row have a probability ...

View answer
X is Uniform $[-\theta,\theta]$ what is the distribution of $Y=\frac{1}{x^{2}}$?
Accepted answer
3 votes

First, $G(0) = 0$. For $y > 0$, $P(x^{-2} \leq y) = P(x^2 \geq y^{-1}) = P(x \geq y^{-0.5}) + P(x \leq - y^{-0.5})$. You just missed one change from $\leq$ to $\geq$ in the middle.

View answer
If x and y are uncorrelated with z, does is follow that x/y uncorrelated with z?
Accepted answer
3 votes

The answer is no, here is a counter example: $P(y = 1) = P(y = -1) = 0.5, ~ x = z y$ with $P(z = 1) = P(z = 2) = 0.5$ and $z$ independent of $y$. Since $z = x/y$, $z$ and $x/y$ are perfectly ...

View answer
Maths Behind Dummy Variable in Linear Regression (One Hot Encoding)?
Accepted answer
2 votes

In your notation $B_2$ describes the difference between the effects of being female and being male. Everything else is in the intercept $B_0$. Consider this example: Assume a imaginary linear ...

View answer
Mahalanobis distance - understanding the formula
Accepted answer
2 votes

Maybe your understanding is helped by rewriting a part of the formula as: $(x-\mu)^\top S^{-1} (x-\mu) = (S^{-1/2} (x -\mu))^\top S^{-1/2} (x-\mu)$ Then you can see that both $x$ and $x^\top$ are ...

View answer
Linear Model with Categorical Predictor Variable: What's the Math behind it?
2 votes

summary(mod) gives you a nice overview over what's going on under the hood. Basicallly what happens is that the categorical variable is automatically turned into a number of binary variables, i.e. is ...

View answer
I want to create a random sample of length n from a normal multivariate distribution
2 votes

Are you allowed to use a package that provides you with independent normal distributed RVs? If not, the Box-Muller method (https://en.wikipedia.org/wiki/Box%E2%80%93Muller_transform) could be used for ...

View answer
What is the difference between multiple regression WITH interaction, and WITHOUT interaction? (* vs + vs " vs : in lm)
Accepted answer
1 votes

This is a model decision you'll have to make. When fitting a linear model, you have to make assumptions about the true relationship between the predicted variable and the predicting variables. The ...

View answer
Minimum and maximum regularization in L0 (pseudo)norm penalized regression
1 votes

To answer the first part of your question: No variable is selected, if it is optimal to not change any element of $\beta$ away from zero, i.e. $y^\top y \leq (y-x_i \hat{\beta_i})^\top (y-x_i \hat{\...

View answer
k-fold cross validation: Force at least m instances in each fold
1 votes

If I understand your question correctly, you have 1300 observations. For some of your outputs, 1295 of 1300 observation have the value zero and 5 obervations have a value $\neq$ 0. Treated ...

View answer
Beta distribution and normalization
1 votes

These 4 plots show the probability densitiy functions of some distributions, therefore they do not need to be in the range of $[0,1]$. They are not cumulative distribution functions.

View answer
How to find correlation between categorical independant variable and a continuous dependent variable?
0 votes

If you don't have too many different values for the categorical variable, you could encode each value as a binary variable (e.g. for colour of car: one variable: 1 if car is green, 0 else). For these ...

View answer
Generate Gaussian process with squared exponential covariance function
0 votes

user20160 is right. Just to add a little technical help: you simulate data from a Gaussian distribution with covariance matrix $K$ by calculating the Cholesky-Decomposition $K = L L^\top$ and ...

View answer
Data level at which the Regression should be run
0 votes

Using a mixed model might be a solution to your problem. It avoids having 400+ dummy variables. A mixed model might also be preferable to dropping the variables Region and Supplier completely, because ...

View answer