Reputation
110,612
Next tag badge:
269/100 score
19/20 answers
Badges
10 194 394
Newest
 Enlightened
Impact
~3.1m people reached

10h
answered $\chi^2$ tabulated value
10h
awarded  Enlightened
10h
awarded  Nice Answer
12h
comment $\chi^2$ tabulated value
The plot simply seems to recapitulate what the OP has already observed! As you write, the question is not what is happening, but why. Sure, some things are obviously changing when the DF is varied: but why should that cause all entries in the table to increase uniformly down each column? (Not all tabulated distributions have this property.) Notice, too, that the table is constructed to support two-tailed tests as well as tests of either tail. Food for thought: historically it would have worked just as well to tabulate $\chi^2_\nu/\nu$--but those values would not always go up.
13h
comment Get important features of n samples
Thank you for the amplification. I'm not sure how anyone could answer it objectively, though, because you don't seem to have conveyed any information we can use to help you quantify "importance." Could you explain what this is intended to mean?
14h
comment cross correlation of partially correlated signals
I do not understand the remarks in (1). How can $c(t)$ be a correlation, given that the notation suggests it could be any function of $t$? That would allow its values to lie beyond the interval of valid correlation values $[-1,1]$.
14h
comment Get important features of n samples
Please edit this post to explain what you are trying to accomplish. What is your analytical objective? Also, by means of the repeated "n" in "n_samples" and "n_features" are you trying to say you have exactly as many observations as you have variables?
14h
comment How can I do full Griffing diallel in Genstat over locations
This question is utterly incomprehensible. I haven't any idea what to recommend, so all I can suggest is that you read the advice in our help center about how to ask questions and then rewrite this post from the beginning.
14h
comment $\chi^2$ tabulated value
Note that the behavior of the median will have little bearing on the question about upper percentiles. You will also find that working with the ratio of the CDFs is difficult: they are normalized incomplete Gamma functions and their derivatives (with respect to the DF parameter) have obscure expressions that most people would find difficult to work with.
14h
comment $\chi^2$ tabulated value
This is the basis of a good explanation but it's logically flawed. The mean could conceivably increase while the tail probabilities decrease. Even when the mean and SD both increase, a decrease in some tail probabilities is possible (and not entirely ruled out by Chebyshev's Inequality). Thus, a subtler argument is needed.
14h
comment Statistical Significance
The phrase "beyond the threshold" suggests you have in mind a particular kind of hypothesis test which is not universally applicable. To be more general you should consider referring to "critical regions" rather than any particular threshold. The reference to a "density curve" also presupposes a narrow application. Many test statistics--especially those involved in non-parametric procedures--have no density at all.
14h
comment Is polynomial regression restricted to linear models?
What you mean by "polynomial regression" is merely a special case of multiple regression. This is explained in detail at stats.stackexchange.com/a/148713.
15h
comment Is it possible (and if yes how) to retain a sparse matrix after normalization?
Re the edit: could you expand on what needs to be "mitigated" and why you would like to do this? I can foresee serious problems. For instance, consider a case where the matrix is binary--all entries are zeros or ones. The procedure you propose (of reducing all nonzero values by the mean of nonzero values) would always reduce any such matrix to all zeros, thereby obliterating all the information in it! This hints at how radical--and potentially meaningless--that proposed operation might be, even when applied to general sparse matrices.
15h
comment $\chi^2$ tabulated value
Although this particular picture is consistent with the table, it doesn't constitute either an explanation or a proof.
16h
comment Exact formula Yates' correction in R
+1 Thank you for making the connection to that last form of the statistic.
16h
revised Continuous Independence Test using only sufficient statistics
edited tags
16h
comment Univariate Linear regression
I'm afraid none of that has any meaning to me (nor will it for many readers of this forum). Could you restate it in English or in mathematical notation? What does "first to last" mean?
16h
comment Significance of regression coefficients and their equality
The crux of the matter is that the variance of $\hat\beta_2-\hat\beta_1$ will typically be greater than either of the variances of $\hat\beta_2$ or $\hat\beta_1$. Thus the width of a CI based on $\hat\beta_2-\hat\beta_1$ could be substantially greater than the widths of CIs based on either estimate alone.
16h
reviewed Leave Open Is it possible (and if yes how) to retain a sparse matrix after normalization?
16h
comment Is it possible (and if yes how) to retain a sparse matrix after normalization?
I think I follow the intent of the question, but it seems to have ineluctably conflicting aims: if the resulting matrix must be sparse, then in what sense can you claim to have "removed" column effects? I can imagine that being possible only by changing the sense of "removed" from "subtracted" to some other operation that will not change the zeros--such as a division. Could you perhaps clarify your sense of "removing" a column effect?