Retrieving "signal" buried far below noise

If some of what follows seems familiar, see 
http://groups.google.com/groups/search?q=%22Time+cards+and+sampling+theorems%22+group%3Acomp.dsp
where I was given "universal advice" , "DON'T" ;)

I drive a delivery route on a *NOMINAL*(LOL) 10 hour 400 mile route with 
1 to 7 stops. Shift can go up >14 hours, mileage can go to ~600 miles. 
Vehicles used can range from minivan (Dodge Caravan) to full size cargo 
van (Dodge Sprinter). Data available would cover >3 drivers and at least 
4 vehicles. This is a 7 day 52 week operation. Holiday and weekend times 
a significantly lower. Available time data lumps driving and 
loading/unloading/reloading times together.

Cruise control has just been added to my primary vehicle. As I expected 
it has reduced my driving time. Don't have enough experience to say just 
how much. I suspect ~15 minutes per day minimum.

Do I have a chance of proving it with available data?
If it is correct, there will be a beneficial cost/benefit ratio to place 
cruise on "required options" for future vehicles.

Any suggested tools for data analysis?

Richard Owlett wrote:
>
> Cruise control has just been added to my primary vehicle. As I expected
> it has reduced my driving time. Don't have enough experience to say just
> how much. I suspect ~15 minutes per day minimum.
>
> Do I have a chance of proving it with available data?
> If it is correct, there will be a beneficial cost/benefit ratio to place
> cruise on "required options" for future vehicles.
>
> Any suggested tools for data analysis?

Hello Richard,

Break your driving data into two sets. One is the driving times without
cruise control and the other is the set of times while using cruise
control. Find the mean and standard deviation of each set. Now assuming
the distributions are gaussian, plot two bell curves - one for each set
of data. Do the curves overlap a lot? A high amount of overlap
indicates either the means are similar or your variance in trip times
is very high so there is a lot of uncertainty or you may have a mix of
both going on. There are some statistical tests you can do on the
difference of means so you can get an idea of the uncertainty that
using a cruise control is advantagous. You can do this in Excel.

IHTH,

Clay

Clay wrote:
> Richard Owlett wrote:
>
> > Cruise control has just been added to my primary vehicle. As I expected
> > it has reduced my driving time. Don't have enough experience to say just
> > how much. I suspect ~15 minutes per day minimum.
>
> > Do I have a chance of proving it with available data?
> > If it is correct, there will be a beneficial cost/benefit ratio to place
> > cruise on "required options" for future vehicles.
>
> > Any suggested tools for data analysis?Hello Richard,
>
> Break your driving data into two sets. One is the driving times without
> cruise control and the other is the set of times while using cruise
> control.

Exactly. And the routes must be the same, otherwise you can't prove
anything! You need measurements of the same route (preferably several,
all with equal vehicles, etc.), once using a vehicle of type A equipped
with cruise control, and once using the same vehicle without cruise
control.

> Find the mean and standard deviation of each set. Now assuming
> the distributions are gaussian, plot two bell curves - one for each set
> of data. Do the curves overlap a lot? A high amount of overlap
> indicates either the means are similar or your variance in trip times
> is very high so there is a lot of uncertainty or you may have a mix of
> both going on. There are some statistical tests you can do on the
> difference of means so you can get an idea of the uncertainty that
> using a cruise control is advantagous. You can do this in Excel.

If the distribution is Gaussian, then one can use t tests. However, I
doubt it is Gaussian, in this case other tests (for example the
Mann-Whitney U-test) are more appropriate. You can use one-sided tests
to show that cruise control decreased the average round-trip time.
Whether you'll achieve significance for the decided niveau depends on
the number of measurements you have available. The more the better.

Showing _how much_ the average time was decreased is another story, but
that might interest managment as well. Starting with graphing the data
is always a good idea.

Regards,
Andor

Andor wrote:
> 
> Clay wrote:
> 
>>Richard Owlett wrote:
>>
>>
>>>Cruise control has just been added to my primary vehicle. As I expected
>>>it has reduced my driving time. Don't have enough experience to say just
>>>how much. I suspect ~15 minutes per day minimum.
>>
>>>Do I have a chance of proving it with available data?
>>>If it is correct, there will be a beneficial cost/benefit ratio to place
>>>cruise on "required options" for future vehicles.
>>
>>>Any suggested tools for data analysis?Hello Richard,
>>
>>Break your driving data into two sets. One is the driving times without
>>cruise control and the other is the set of times while using cruise
>>control.
> 
> 
> Exactly. And the routes must be the same, otherwise you can't prove
> anything! You need measurements of the same route (preferably several,
> all with equal vehicles, etc.), once using a vehicle of type A equipped
> with cruise control, and once using the same vehicle without cruise
> control.
> 
> 
>>Find the mean and standard deviation of each set. Now assuming
>>the distributions are gaussian, plot two bell curves - one for each set
>>of data. Do the curves overlap a lot? A high amount of overlap
>>indicates either the means are similar or your variance in trip times
>>is very high so there is a lot of uncertainty or you may have a mix of
>>both going on. There are some statistical tests you can do on the
>>difference of means so you can get an idea of the uncertainty that
>>using a cruise control is advantagous. You can do this in Excel.
> 
> 
> If the distribution is Gaussian, then one can use t tests. However, I
> doubt it is Gaussian, in this case other tests (for example the
> Mann-Whitney U-test) are more appropriate. You can use one-sided tests
> to show that cruise control decreased the average round-trip time.
> Whether you'll achieve significance for the decided niveau depends on
> the number of measurements you have available. The more the better.
> 
> Showing _how much_ the average time was decreased is another story, but
> that might interest managment as well. Starting with graphing the data
> is always a good idea.
> 
> Regards,
> Andor
> 

I was afraid I didn't have a chance. Driving times would be strongly 
correlated with day of week - eg weekends. Loading/unloading times would 
also be strongly correlated with DOW -- but with "weekend effect" 
shifted by one day. And the time data would be total clockin to clockout.

Thanks.

Richard Owlett wrote:
> Andor wrote:
>
> > Clay wrote:
>
> >>Richard Owlett wrote:
>
> >>>Cruise control has just been added to my primary vehicle. As I expected
> >>>it has reduced my driving time. Don't have enough experience to say just
> >>>how much. I suspect ~15 minutes per day minimum.
>
> >>>Do I have a chance of proving it with available data?
> >>>If it is correct, there will be a beneficial cost/benefit ratio to place
> >>>cruise on "required options" for future vehicles.
>
> >>>Any suggested tools for data analysis?Hello Richard,
>
> >>Break your driving data into two sets. One is the driving times without
> >>cruise control and the other is the set of times while using cruise
> >>control.
>
> > Exactly. And the routes must be the same, otherwise you can't prove
> > anything! You need measurements of the same route (preferably several,
> > all with equal vehicles, etc.), once using a vehicle of type A equipped
> > with cruise control, and once using the same vehicle without cruise
> > control.
>
> >>Find the mean and standard deviation of each set. Now assuming
> >>the distributions are gaussian, plot two bell curves - one for each set
> >>of data. Do the curves overlap a lot? A high amount of overlap
> >>indicates either the means are similar or your variance in trip times
> >>is very high so there is a lot of uncertainty or you may have a mix of
> >>both going on. There are some statistical tests you can do on the
> >>difference of means so you can get an idea of the uncertainty that
> >>using a cruise control is advantagous. You can do this in Excel.
>
> > If the distribution is Gaussian, then one can use t tests. However, I
> > doubt it is Gaussian, in this case other tests (for example the
> > Mann-Whitney U-test) are more appropriate. You can use one-sided tests
> > to show that cruise control decreased the average round-trip time.
> > Whether you'll achieve significance for the decided niveau depends on
> > the number of measurements you have available. The more the better.
>
> > Showing _how much_ the average time was decreased is another story, but
> > that might interest managment as well. Starting with graphing the data
> > is always a good idea.
>
> > Regards,
> > AndorI was afraid I didn't have a chance. Driving times would be strongly
> correlated with day of week - eg weekends. Loading/unloading times would
> also be strongly correlated with DOW -- but with "weekend effect"
> shifted by one day. And the time data would be total clockin to clockout.

One could account for the week day effect, as long as you have data
from the same route using the same vehicle on the same weekday ...

Andor wrote:

   ...

> Whether you'll achieve significance for the decided niveau depends on
> the number of measurements you have available. The more the better.

   ...

"Niveau" means "level" in French, adopted into German. What does it mean 
in English?

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Jerry Avins wrote:
> Andor wrote:   ...
>
> > Whether you'll achieve significance for the decided niveau depends on
> > the number of measurements you have available. The more the better.   ...
>
> "Niveau" means "level" in French, adopted into German. What does it mean
> in English?

Level.

:-)

Sorry, sometimes German or French words slip into my English.

Andor wrote:
> 
> Jerry Avins wrote:
>> Andor wrote:   ...
>>
>>> Whether you'll achieve significance for the decided niveau depends on
>>> the number of measurements you have available. The more the better.   ...
>> "Niveau" means "level" in French, adopted into German. What does it mean
>> in English?
> 
> Level.
> 
> :-)
> 
> Sorry, sometimes German or French words slip into my English.

Sei Gesunt!

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

well,
       this technique of retrieving signal buried under noise is quite
common in acoustic signals..in underwater navigation..There is wavelet
based Denoising techniques..on which you can do denoising and then you
have to do DOA estimation..

so..whoever has posted this question..do also look at WAVELET BASED
DENOISING TECHNIQUES..which work quite well..for above scenarios stated

particlereddy


Jerry Avins wrote:
> Andor wrote:
> >
> > Jerry Avins wrote:
> >> Andor wrote:   ...
> >>
> >>> Whether you'll achieve significance for the decided niveau depends on
> >>> the number of measurements you have available. The more the better.  =
 =2E..
> >> "Niveau" means "level" in French, adopted into German. What does it me=
an
> >> in English?
> >
> > Level.
> >
> > :-)
> >
> > Sorry, sometimes German or French words slip into my English.
>
> Sei Gesunt!
>
> Jerry
> --
> Engineering is the art of making what you want from things you can get.
> =AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=
=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF=AF

I'm in Springfield MO under ice storm. Have not had power for >24 hours.
Suspect will not have reliable power for 2-3 days. Have heat and shelter 
-- what else do i need.

Will reply when normalcy returns.

Thank you as by backup battery decayyyyyys