10,810 reputation
11128
bio website
location
age 51
visits member for 3 years, 10 months
seen 2 hours ago

Nov
28
comment Using kFold in a regression problem?
It rather depends on what you mean by regression. $k$-fold cross validation is method to help decide the hyperparameters of your method (at an extreme level deciding the method itself) while trying to avoid over-fitting. And you should not be using the test set to compare methods: you should instead be using the results of the cross validation.
Nov
23
comment Area within a given number of standard deviations from given mean
It depends on the shape of the distribution.
Nov
23
comment Rearrange boxplot with ggplot
It may depend on whether you use reshape or reshape2 (you should have mentioned this in your question). As for displaying results, you can show how many 1s and 0s there are or what proportion are 1s.
Nov
19
comment Fast way to compute central moments of a Poisson random variable?
What do you know? The moment generating function? Stirling numbers of the second kind? Moving from moments about $0$ to moments about the mean?
Nov
18
comment Can mean plus one standard deviation exceed maximum value?
Indeed - my minor point is that this curiosity is a result of what standard deviations represent for strongly non-symmetric distributions rather than a result of taking a sample. But in general, I think your answer is excellent
Nov
18
comment Can mean plus one standard deviation exceed maximum value?
Pedantically, you and R are using the $n-1$ sample standard deviation calculation. If the population is $1,5,5,5$ then its standard deviation is $\sqrt{3} \gt 1$ so your example is still valid.
Nov
16
comment Inconsistency between R and SAS for MLE on Weibull
The difference between the two sets of parameters is much smaller than the bracketed estimates of the errors in these estimates. So they are close.
Nov
16
comment Polynomial regression P value is getting altered
Almost certainly yes if you are going to use the raw polynomials in the final equation, probably yes even if not. And the 1st (linear) and 0th (constant) degrees too.
Nov
16
comment Polynomial regression P value is getting altered
I am not a fan of polynomial regression without a theoretical justification, though it might sometimes work for interpolation (not extrapolation) as a way of drawing a smooth curve. If you must go down this route, you might consider using the orthogonal polynomials to decide which is the highest degree you are going to take into account, and then re-regress on the raw polynomials of that degree and smaller to get coefficients which can easily be implemented, even though the reported statistics about the coefficients of the re-regressed raw polynomials are meaningless.
Nov
11
comment Creating condifence intervals ? Help with statistics intervals.
So which are you finding difficult in the calculation of $\bar{x} + z s / \sqrt{n}$? A sample of $50$ should be big enough to assume a normal distribution is a reasonable approximation of the distribution of the sample mean. If you would rather use a $t$-distrubution, see an earlier answer. (Your use of $\mu$ is not quite correct, but that should not stop you calculating the confidence interval)
Nov
11
comment null hypothesis confusion
Truth? Or what the person asking the question wants you to say, namely that the null hypothesis is what the experts argue, i.e. that $\le 30\%$ of schools pupils benefit from homework? And the alternative hypothesis is that more do?
Nov
9
comment sample size power calculation
The power of a test is the probability that it correctly rejects the null hypothesis when the null hypothesis is false. What is your null hypothesis? If it is "This individual does not have the disease" then sample size is irrelevant as you already know the sensitivity. If it some statement about the group, then it would useful to know what that is.
Nov
7
comment How to estimate model where instrument is correlated with dependent variable
It makes sense, but note that for example errors in estimating $b_1$ will lead to correlated errors in estimating $\beta_4$ and $\beta_5$ and so you will be less confident about the estimates of each of them.
Nov
2
comment How to solve this problem on Curse of Dimensionality problem - Nearest Neighbours
If you are using a Euclidean metric then it would be the hypersphere.
Oct
29
comment Is a confidence interval a prediction interval for the sample mean?
See stats.stackexchange.com/questions/16493/… for more
Oct
29
comment Is a confidence interval a prediction interval for the sample mean?
That is not the standard definition of a $95\%$ confidence interval, which is more along the lines of applying the same methodology to repeated independent sample means with the same sample size, $95\%$ of the confidence intervals so constructed would include the population mean. I also have doubts about your description of the prediction interval, which does not seem to depend on the actual value of the sample mean observed. A difference is that the prediction interval takes account of the uncertainty of the next sample mean compared with the population mean, in addition to this sample's.
Oct
29
comment What Ratio of Independent Distributions gives a Normal Distribution?
A rather special case is $X$ a standard normal, and $Y$ independently $\pm1$ each with probability $\frac12$, then $X$, $Y$ and $\frac{X}{Y}$ have the same mean and variance and $\frac{X}{Y}$ is normally distributed.
Oct
27
comment What is a log-odds distribution?
It is possible that your model could produce arbitrarily large positive or negative results. These might not be interpretable in terms of a bounded range such as a probability, but could be interpretable as log-odds using the logit function and its inverse the logistic function.
Oct
18
comment Visualize a large binary matrix with instances and three classes
You can visualise an matrix for example using image but you might want to reorder the rows and columns first into clusters or something similar
Oct
18
comment How to quantify the importance of a win in soccer?
You would want to give it a rating similar to teams it was perceived to be as likely to win or lose against.