Bayesian stats: how to update probability?

[ireland01]

I am trying to use Bayesian methods (Bayes rule) to predict further datapoints (at point n,n+1,n+2 etc..)...

I begin by generating a normal pdf using previous 75 datapoints (prior: n-75 to n-1) with mean value, μ: 1.25 and standard deviation, δ: 3.67.

Note: previous datapoints range from -5 to +5 in value.

I calculate the maximum probabilty of 0.11 for datapoint = 1.30.

Using this will underestimate (predict) the value at n.

I now want to incorporate (into the probability) the fact that I know my previous two datapoints (n-2 to n-1) showed increase towards +ve...

 

[mfb]

You can give your datapoints nearby a higher weight - there are many ways to do this, depending on the type of correlation between the points you expect.

 

[Stephen Tashi]

I am trying to use Bayesian methods (Bayes rule) to predict further datapoints (at point n,n+1,n+2 etc..)...

.

You'll have to explain what probability model you are using and what prior distributions you are using before anyone can give you an answer. A Bayesian method has to be more than the willingness to update estimates. You must also be willing to assume a specific probability model and specific priors.

If this is a real world problem, describe it and someone might suggest a probability model.

 

[Number Nine]

The probability that your datapoint = 1.3 is zero.
This seems like a very strange thing to do. Are your data dependent in some way? If you're just sampling from a normal distributions, then your previous samples are irrelevant. Calculate your probability directly from the normal pdf.