Re: p-value issue

On 11 Dec 2006 08:25:28 -0800, "Ethan.Johnsons@xxxxxxxxx"
<Ethan.Johnsons@xxxxxxxxx> wrote:


Richard Ulrich wrote:
On 10 Dec 2006 12:55:43 -0800, "Ethan.Johnsons@xxxxxxxxx"
<Ethan.Johnsons@xxxxxxxxx> wrote:

I have these two data sets and plotted them one by one, compare, and
confirm that both return the same curve and shape, close to identical.


x=(3.05176E-05,0.000457764,0.003204346,0.0138855,0.04165649,0.09164429,
0.1527405,0.1963806,0.1963806,0.1527405,0.09164429,0.04165649,0.0138855,
0.003204346,0.000457764,3.05176E-05)
y=(0.0000306,0.0004566,0.0031985,0.0139083,0.0415539,0.0917678,
0.1528134,0.1962831,0.1962994,0.1527996,0.0917336,0.0415497,
0.0139308,0.0031917,0.0004529,0.0000301)

This makes me think that the p-value must be very significant (less
0.05), but if I run Chi-Square, I get p-value = 0.3021.

Should I use a different test to see if they are "identical" or
not?

This post shows a lot of confusion --

1) Which "chi-square" test? The contingency table test is applied
to *counts*, preferably counts greater than 5.0; and all of
your numbers are less than 1. Whatever test you use, there will
need to be some reference to sample size, either in computing
the test or in looking up a p-level.

2) If a Goodness of Fit test has a big chi-square, that says that
the distributions are *different*. So, assuming that you have
used some test that is appropriate, your result says what you
hoped -- your sample size is not large enough to make their
difference 'statistically significant.'

--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html

I used Pearson's Chi-squared test.
I also used t-test, etc, but the p-value is very high. With these data
sets, I want to compare the similarity, which it is very close to
identical if you plot them. I thought that sample size 9 would be
enough to compare two data sets. That is why I was wondering if there
is another test to confirm the similarity.

Pearson's contingency chi-squared, or Pearson's used for
Goodness of Fit, uses integer counts in the cells.
Your data is something from somewhere else. Does it
represent tens of thousands in total, in order to have those
tiny fractions?

You apparently need to go back to the start, and learn what
these tests do, and something about how they do it.
Pick up a textbook or two and browse.

--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.

Relevant Pages

  • Re: p-value issue
    ... Richard Ulrich wrote: ... This makes me think that the p-value must be very significant (less ... 0.05), but if I run Chi-Square, I get p-value = 0.3021. ... is another test to confirm the similarity. ...
    (sci.stat.math)
  • Re: p-value issue
    ... Richard Ulrich wrote: ... This makes me think that the p-value must be very significant (less ... 0.05), but if I run Chi-Square, I get p-value = 0.3021. ... is another test to confirm the similarity. ...
    (sci.stat.math)
  • Re: p-value issue
    ... 0.05), but if I run Chi-Square, I get p-value = 0.3021. ... I also used t-test, etc, but the p-value is very high. ... is another test to confirm the similarity. ... The plot however lets assume a quadratically increasing distance between points - or/and a time series. ...
    (sci.stat.math)
  • Re: Variable Importance
    ... A chi-square with more than 1 d.f. ... Given two factors with the same p-value ... you need to look at both the univariate tests ... the validity of measurements. ...
    (sci.stat.consult)
  • Re: p-value issue
    ... confirm that both return the same curve and shape, ... This makes me think that the p-value must be very significant (less ... 0.05), but if I run Chi-Square, I get p-value = 0.3021. ... the distributions are *different*. ...
    (sci.stat.math)
Loading