Hi all, Thanks for the responses, I will try and explain the problem more clearly using a different example: Lets assume we have a room filled with sensors (ex. temperature sensor, a TV sensor that senses if the TV is on or not, a weight sensor on the floor, a pressure sensor on the tables, ...etc.). Lets assume also that activities are taking place in that room (e.g. dinner, lunch, watching a movie, chatting on the phone,...etc.). Assume for the sake of this example, we have 4 activities, the activities can possibly be detected using particular sensors (Ex. watching the TV can be detected using the TV sesnor, maybe the weight sensor of the couch infront of the TV (i.e. someone is watching), possibly also a camera image that determines whether the person is looking at the screen or not). My goal is to determine what sensor set to use for infering the activities. The data set set collected from the space has sensor readings along with the activities. Moreover, I want to construct a bayesian network for every activity. Notice that if I include sensors that are somehow linearly dependant, the bayesian network will get multiple evidence from the same source, so thats why I would like to eliminate highly correlated variables. The approach that I am using to tackle this problem, is : 1) Variable screening using logistic regression where the response variable is the activity (non-ordinal, categorical) and the sensors are the independent variables. 2) Building a bayesian model using the variables selected 3) determining the probabilities and likelihoods for the bayesion model using the collected data
Does that make sense? Fahd "Paige Miller" <[EMAIL PROTECTED]> wrote in message news:[EMAIL PROTECTED] > albinali wrote: > > Hi, > > I am using a set of data that I collected using biosensors, the goal is to > > analyze the data > > Okay, let's stop right there. We all try to analyze data. What > specifically does "analyze the data" mean in your application? > > > and to infer the dependencies between the variables to > > More problem definition would be good here, not in statistical terms, > but in terms of your application. And what "variables" do you mean? If > you mean predictor variables (the biosensors, I assume), that's one > thing; if you mean between the biosensors and the mood (mentioned below) > then that's a different thing altogether. > > > construct a bayesian network that infers the mood of a person (e.g. happy, > > sad...etc). > > Oh, why didn't you say so? This is a very easy problem. You can > determine the person's mood by the little smiley face in his e-mail. :-) > > Okay, I'm being facetious here, back to serious now. > > > Clearly, the response variable is non-ordinal and categorical. > > I'm not sure about non-ordinal, but it certainly seems to be > categorical. Is it univariate or multivariate? > > > The model should not include interdependent sensors. > > There are only two ways to get non-interdependent sensors, if I am > understanding you properly. > > One way is to design an orthogonal experiment, and since you are talking > about biosensors, I don't think you can do that here. The second way is > to take the sensors you have and find orthogonal combinations of them > (in effect, computing new sensors) by something like Principal > Components Analysis (PCA) or Partial Least Squares (PLS). These two > methods are also useful if I understand your earlier stated need "to > infer the dependencies between the variables". > > > I would like to get > > some thoughts about using regression for such a task, is it a good choice ? > > "Regression" is a broad term. In one sense of the word, regression can > include any predictive method, such as logistic regression, neural > networks and the aforementioned Bayesian networks. Any of those might be > good candidates for your case, which does not have a continuous response > variable. Ordinary Least Squares, which is a common use of the term > "regression", does not seem to fit because your response is not continuous. > > > are there any other possible techniques to use? > > I think I have mentioned a few. > > > More importantly, would it > > be possible to automate the building of such models? > > Yes, but this can be very dangerous. > > > Is anyone aware of any earlier work in that area? > > I'd bet the answer is YES to this question too. > > -- > Paige Miller > Eastman Kodak Company > [EMAIL PROTECTED] > http://www.kodak.com > > "It's nothing until I call it!" -- Bill Klem, NL Umpire > "When you get the choice to sit it out or dance, I hope you dance" -- > Lee Ann Womack > . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
