| |
 |
|
| Science Forum Index » Space - Consult Forum » How to check for similarity in the results...help!... |
|
Page 1 of 1 |
|
| Author |
Message |
| hall... |
 Posted: Wed Sep 16, 2009 8:31 am |
|
|
|
Guest
|
Hi,
I have the precision values of a system on two different data sets.
The snippets of these results are as shown:
sample1: (total 194 samples)
0.6000000238
0.8000000119
0.6000000238
0.2000000030
0.6000000238
....
....
sample2: (total 188 samples)
0.80000001
0.20000000
0.80000001
0.00000000
0.80000001
0.40000001
....
....
I want to check how much statistically significant these results are?
Is unpaired t-test the right test to do?
Nevertheless, i performed t-test on the data and got the following
result:
-------------------------------------------------------
Welch Two Sample t-test
data: s1p5 and s2p5
t = 0.9778, df = 374.904, p-value = 0.3288
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.03170059 0.09441172
sample estimates:
mean of x mean of y
0.5138298 0.4824742
-------------------------------------------------------
I am not able to interpret these results. I believe p value less than
0.05 means, the results are very different but how to check for
similarity in the results? Does higher p-value while testing for the
dis-similarity means results are similar?
please help. |
|
|
| Back to top |
|
|
|
| Rich Ulrich... |
| Posted: Sun Sep 20, 2009 9:35 pm |
|
|
|
Guest
|
On Sat, 19 Sep 2009 06:24:07 -0700 (PDT), hall
<hall.robert.anu at (no spam) gmail.com> wrote:
[quote:a9fe4eae73]Can i merely comapre the means and variances of the lists and if they
are 'similar' then conclude that the two lists are similar?
On Sep 18, 1:25 am, Rich Ulrich <rich.ulr... at (no spam) comcast.net> wrote:
On Wed, 16 Sep 2009 11:31:10 -0700 (PDT), hall
[snip, extensive previous][/quote:a9fe4eae73]
Sorry, I think I got carried away by your inappropriate
listing of the data to 10 or so decimals, and I wrongly assumed
you were asking some interesting question about "precision".
- If your data are only useful to two or three places, you should
generally present only that much precision.
If you just want to know that the means are similar, you
present the test and the CI for the means, as you did,
and, perhaps, test the variances.
--
Rich Ulrich |
|
|
| Back to top |
|
|
|
| hall... |
| Posted: Sun Sep 20, 2009 11:44 pm |
|
|
|
Guest
|
On Sep 21, 8:35 am, Rich Ulrich <rich.ulr... at (no spam) comcast.net> wrote:
[quote:79b0e15739]On Sat, 19 Sep 2009 06:24:07 -0700 (PDT), hall
hall.robert.... at (no spam) gmail.com> wrote:
Can i merely comapre the means and variances of the lists and if they
are 'similar' then conclude that the two lists are similar?
On Sep 18, 1:25 am, Rich Ulrich <rich.ulr... at (no spam) comcast.net> wrote:
On Wed, 16 Sep 2009 11:31:10 -0700 (PDT), hall
[snip, extensive previous]
Sorry, I think I got carried away by your inappropriate
listing of the data to 10 or so decimals, and I wrongly assumed
you were asking some interesting question about "precision".
- If your data are only useful to two or three places, you should
generally present only that much precision.
If you just want to know that the means are similar, you
present the test and the CI for the means, as you did,
and, perhaps, test the variances.
--
Rich Ulrich
[/quote:79b0e15739]
I also want to know what these results mean? :-
the samples of s1p5 and s2p5 mentioned above in the thread.
--------------------------------------------------
Welch Two Sample t-test
data: s1p5 and s2p5
t = 0.9778, df = 374.904, p-value = 0.3288
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.03170059 0.09441172
sample estimates:
mean of x mean of y
0.5138298 0.4824742
-------------------------------------------------
what does such a large 'p-value' mean?
Please somebody interpret these results.
at (no spam) Rich : thanks  |
|
|
| Back to top |
|
|
|
| Rich Ulrich... |
| Posted: Tue Sep 22, 2009 1:06 pm |
|
|
|
Guest
|
On Mon, 21 Sep 2009 02:44:12 -0700 (PDT), hall
<hall.robert.anu at (no spam) gmail.com> wrote:
[quote:471df61f22]On Sep 21, 8:35 am, Rich Ulrich <rich.ulr... at (no spam) comcast.net> wrote:
On Sat, 19 Sep 2009 06:24:07 -0700 (PDT), hall
hall.robert.... at (no spam) gmail.com> wrote:
Can i merely comapre the means and variances of the lists and if they
are 'similar' then conclude that the two lists are similar?
On Sep 18, 1:25 am, Rich Ulrich <rich.ulr... at (no spam) comcast.net> wrote:
On Wed, 16 Sep 2009 11:31:10 -0700 (PDT), hall
[snip, extensive previous]
Sorry, I think I got carried away by your inappropriate
listing of the data to 10 or so decimals, and I wrongly assumed
you were asking some interesting question about "precision".
- If your data are only useful to two or three places, you should
generally present only that much precision.
If you just want to know that the means are similar, you
present the test and the CI for the means, as you did,
and, perhaps, test the variances.
--
Rich Ulrich
I also want to know what these results mean? :-
the samples of s1p5 and s2p5 mentioned above in the thread.
--------------------------------------------------
Welch Two Sample t-test
data: s1p5 and s2p5
t = 0.9778, df = 374.904, p-value = 0.3288
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.03170059 0.09441172
sample estimates:
mean of x mean of y
0.5138298 0.4824742
-------------------------------------------------
what does such a large 'p-value' mean?
Please somebody interpret these results.
at (no spam) Rich : thanks
[/quote:471df61f22]
A p-value of 1/3 is pretty mediocre. It says that
if the groups have underlying means that are the same,
you would see a difference this large at least 1/3 of the
time.
For "interpreting the results", the confidence interval gives
some further guidance. That gives the range which is most
likely to include the true difference. Are the means
close enough for your purposes? Or, Is the difference
small enough for your purposes?
You conclude that the sample is not large enough to
show that the means are different, if indeed they are
different. One characterization of the likely range of
difference is shown by that 95% CI.
--
Rich Ulrich |
|
|
| Back to top |
|
|
|
|
|
All times are GMT - 5 Hours
The time now is Mon Nov 23, 2009 5:18 am
|
|