Ask Experts Questions for FREE Help !
    janguera's Avatar
    janguera Posts: 7, Reputation: 1
    New Member

    Sep 30, 2005, 11:55 AM
    Exponential statistical contrast
    I'm trying to figure out this problem: iI'm given a maximum and minimum value for 11 consecuitive values, and these values when plotted fit an exponential curve. So for example the values are:

    I'm trying to set up an exponential statistical contrast here which basically means that I'm trying to assign values to each of the aforementioned data points that when plotted would decay in an exponential function at a similar rate/slope as the data plotted above (same trendline equation) AND if these weighted values are summed together they add up to zero. I know that it is importatnt to consider tau (where the exponential curve decays to ~37% its original value) for my values, but I'm getting stuck assigning weighted values.

    Any help would be greatly appreciated!

    joseph2000's Avatar
    joseph2000 Posts: 12, Reputation: 1
    New Member

    Oct 1, 2005, 06:34 AM
    You can use the what's best modelling of the lindo product. This is an add-in for ms excel. You can point to sum of the reference cell equal to zero... and the values will be generated as long as you defined its correct slopes and other parameters.

    janguera's Avatar
    janguera Posts: 7, Reputation: 1
    New Member

    Oct 1, 2005, 12:49 PM
    Thanks, I've been playing with What's best for the afternoon learning how to use it and now have a couple of questions so that I'm not here all night!

    -I'm guessing that I'd have to put in a maximum and minimum values...

    -Likewise, the shape of the curve would be based on the equation
    y = 23.855e-0.1308x for the values below... I'm just not sure how to set this up next... if you (or anybody) is knowledgeable regarding this program, please let me know!

    joseph2000's Avatar
    joseph2000 Posts: 12, Reputation: 1
    New Member

    Oct 2, 2005, 03:56 AM
    Use of what is best

    IF what you want is a parallel line relative to the given line then the slope must be equal to this line. You can use the required values using this for the coordinates of points. May I see your problem so that I can play with it too. YOu may also use solver found in add-in of excel, though it is less powerful.

    janguera's Avatar
    janguera Posts: 7, Reputation: 1
    New Member

    Oct 2, 2005, 08:40 AM
    Hello Again!
    I'd be happy to describe the problem more in-depth if it leads to an answer, as my trying to model it in What's Best has been futile to this point!

    So given the (11) values that I first posted, when I fit an exponential trendline to these points (after graphing them in EXCEL), I get the equation of
    y = 23.855e-0.1308x. So what I need to do now is come up with arbitrary weighted values for these points that if I were to plot them would:

    a) give me the same exponentially degrading shape as my original data points if they were plotted.
    b) these weighted values when summed together add up to zero. So the first data point should be given 'more weight' than the second, the second 'more' than the third, etc.
    Here's an example of what these weights would look like:


    The aforementioned weights were assigned to the following data set:


    I don't know what the equation for the trendline was in this case, but maybe this knowledge will help me help you! Thanks again, I hope that this makes the question that much more clear!

    joseph2000's Avatar
    joseph2000 Posts: 12, Reputation: 1
    New Member

    Oct 3, 2005, 07:36 AM
    initial solution
    Hi I hope this is the solution. The answer is:


    I solve first the slope of the given line. I used the x component starting from 11 to 1. By basing on the slope of the line I used the goal seek feature of ms excel. I need to find the correct intercept which will give the sum to be equal to zero.

    There is something wrong with this:
    1, the assumption is that the line will be straight.
    2. I had not considered the residual.
    3. I had not considered what other trend is possible.

    I can do some refinements when I test it with other model that will give the best fit.

    Joseph 2000

Not your question? Ask your question View similar questions


Question Tools Search this Question
Search this Question:

Advanced Search

Add your answer here.

Check out some similar questions!

Statistical method to use [ 1 Answers ]

Hi, I am trying to figure out which method to use to find steady state. So I have data that was collected at 38 time points for 12 individuals. I am supposed to find the time point at which steady state was reached. I want to know what is the best method to answer this question. I have tried to...

SAS(statistical analysis system) [ 1 Answers ]

Hi, I've got some issues pertaining to the SAS software and BaseSAS. I'm trying to locate some links or forums to post my issues. Can anyone help me in finding some... Thanks in advance... SAI

View more questions Search