Hi guys,
Here's a simple question for you math nerds. It'll probably take you 5
mins to get this done. If you answer it exactly as I desire, I will
offer you even upto $20.00.
I have linked an excel file here for you. You may need to have a yahoo
id for the purpose..
http://us.f1f.yahoofs.com/users/f16a9db6/bc/GOOGLE+HELP/Query.xls?bfIgnS_AFIdqmx1J
Within here, you will see different columns of data.
In the file, you will see 5 groups A, B, C, D and E. All you have to
do is to prove by any approved statistical method of analysis that
group A3 is more significant than group A2 and group A1. Similarly for
B, C, D and E.
I dont need any graphs or any pictures. Just write your mode of
analysis with results here in 4-5 sentence statements (you have to
show p values or confidence limits or whatever you use).
All you have to remember is that, it has to be scientifically
statistically SIGNIFICANT.
Cheers,
The crazy mathematician |
Clarification of Question by
crazymathematician-ga
on
25 Aug 2003 21:19 PDT
I APOLOGIZE.
THE FILE CAN BE DOWNLOADED AT http://www.geocities.com/drsojivarghese/Query.xls
Thank you.
|
Request for Question Clarification by
elmarto-ga
on
26 Aug 2003 07:05 PDT
Hi Crazy Mathematician!
Being a math nerd, this question attracted my attention immediately.
However, in order to answer it, I would need to know what do you mean
by a group being "more significant" than some other group. The
significance level is usually applied to "measuring" the veracity of
some hypothesis. What's the hypothesis you're trying to test?
Significane Level
http://davidmlane.com/hyperstat/A72117.html
From the data given, it's clear that all groups labelled "3" have a
greater variance than 1s and 2s; thus they could potentially have a
greater explanatory power... but to explain what? If you could please
clarify this, I'll do my best to answer your question.
Best regards,
elmarto
|
Clarification of Question by
crazymathematician-ga
on
26 Aug 2003 08:10 PDT
Thanks elmarto,
I've checked your link. Perhaps it may be better if I explain what I'm
trying to do.
I need to present this data at a meeting. But I cannot present it
unless the group 3 was more statistically significant that 2 or 1.
Like the link says, "If the probability is less than or equal to the
significance level, then the null hypothesis is rejected and the
outcome is said to be statistically significant. Experimenters have
used either the .05 level (sometimes called the 5% level) or the .01
level (1% level) although the choice of levels is largely subjective.
"
I did a student's t test and did not get even 0.5 level significance
for all the A to E groups. I got it for a couple, but not for all the
groups.
.05 level is fine for me. In fact, you can choose any mode of
analysis, as long as you can let me know in abt 3-4 (or more)
sentences, what to state so that I can present the significance
approvably. All I want the reviewers when they read the 3-4 sentences
is to agree that group 3 is definitely more statistically significant
that group 1 or 2.
Thanks
|
Clarification of Question by
crazymathematician-ga
on
26 Aug 2003 08:22 PDT
Just in case you'd like to know more abt the experiment..
I have 3 groups of people. Group 1, Group 2 and Group 3. My hypothesis
is that Group 3 has more levels of CO2 than group 1 or group 2.
Everything I have done has proved my hypothesis.
I need to now present this data too to that effect.
1 mode of analysis that I have done that shows this is:
If you try comparing the average from the 3 groups, Group3 definitely
has more levels that the average of gp1/2 PLUS 3 Standard Deviations.
That is one way people present to show that Group3 is more
statistically significant that group 2 and 1.
But then again, I dont know how to write that up well to be approved
by the reviewers without any arguments.
Thanks..
|
Request for Question Clarification by
elmarto-ga
on
26 Aug 2003 09:04 PDT
Hi again!
I think I misstated my clarification request. What I meant to say is
that you need an hypothesis you want to test in order to talk about
the significance level. An example of this could be the following:
imagine you are a car manufacturer and that you have three sets of
data: sales, advertisment spending and rain level. Using a regression,
one could scientifically conclude that the level of spending in
advertisment significantly affects sales, and that rain does not have
a significant impact on car sales (with confidence intervals,
p-values, etc). Notice that in order to say that some variable is
significant, one must first ask some question about it ("Does the rain
level affect car sales?", "Is it true that every dollar spent in
advertisment generates an extra two dollars in sales?"). That kind of
question is the one I was asking in my clarification request. If you
can tell me what are you trying to show with this data, maybe I can
assist you with an answer.
Best wishes!
elmarto
|
Request for Question Clarification by
elmarto-ga
on
26 Aug 2003 09:06 PDT
OK, ignore my last clarification request :-) I'll try to address your question now.
Best regards,
elmarto
|
Clarification of Question by
crazymathematician-ga
on
26 Aug 2003 11:11 PDT
Thanks, elmarto. I really appreciate this.
|
Request for Question Clarification by
elmarto-ga
on
26 Aug 2003 15:47 PDT
A usual way to compare means of more than 2 groups is an ANOVA
analysis. With this kind of analysis, it's possible to test the
hypothesis that all groups have equal mean versus the hypothesis that
some pair of groups have different means. I tested this hypothesis
using this methodology with a statistics software (EViews) and found
the same as you did: for some of the letters (A, for example), the 3
means are not significantly different; possibly due to the large
variance of group 3 and the small number of observations for the other
2 groups. Is it possible for you to get more observations for groups 1
and 2? Are you sure that group 3 should have a larger mean than the
other 2 groups?
Best regards,
elmarto
|
Clarification of Question by
crazymathematician-ga
on
26 Aug 2003 19:44 PDT
Thanks, elmarto.
Unfortunately, the project has come to a close and further datapoints
cannot be obtained at this time. So further observations cannot be
obtained at this time.
Yes, group 3 should have a larger mean.
In fact, like I stated earlier, if you try comparing the average from
the 3 groups, Group 3 definitely has on an average higher levels that
the average of group1 or group2 PLUS THREE standard deviations.
I've seen people present this way before and I do know that this is an
acceptable mode of data presentation. But I dont know how to
scientifically state my results with this mode of analysis. If no
other modes of analysis you know can prove my hypothesis, then if you
can write down in 3-4 sentences (or more) how to scientifically state
these results based on that analysis, it will suffice and I will be
happy.
Pls do let me know asap. (sorry- i dont mean to rush you, but i have
to take this in asap..)
thanks a BUNCH
|
Request for Question Clarification by
elmarto-ga
on
27 Aug 2003 08:34 PDT
Hi crazymathematician!
The problem here is that the mean of group 3 is actually not three
standard deviations away from the mean of groups 1 and 2 for all
letters. Take for example group A. The mean of group 3 is 12.66, while
it's standard deviation is 18.39. Results for the other letters show
the same. I'm sorry, but I can't think of any scientific way to prove
what you want to prove with these numbers. Maybe I'm missong
something, but how did you arrive to the fact that the mean of group 3
is more than 3 standard deviations away from the other groups?
In any case, this question is also open to any other Researcher who
might have a better idea to prove it.
Best regards,
elmarto
|
Clarification of Question by
crazymathematician-ga
on
27 Aug 2003 09:51 PDT
Thanks elmarto,
What I was stating was that the MEAN of Group 3 was GREATER than THE
MEAN OF GROUP 1 or GROUP 2 PLUS THREE times their Standard deviation.
SO in this case, in A, group 3 mean which is 19.32 is greater than
0.48 (which is the mean (=0.12) plus 3SD (=0.36)).
Do you see my flow?
I've seen people present like this before.. even on google searches..
But I just dont know how to scientifically state and present this
data. I am not a statistician or a mathematician.
Pls help.
|
Clarification of Question by
crazymathematician-ga
on
27 Aug 2003 10:00 PDT
Hey elmarto,
The example I used was for B. For A, it will be 12.66 (group 3)
greater than 5.04 (against group 2) and 4.2 (against group 1). I've
uploaded the file where it details so.
Pls help.
http://www.geocities.com/drsojivarghese/Query2.xls
|
