Hey aivoryuk.
Can you explain in non-mathematical terms what you are trying to do?
I have been using a simple Logistic Regression method with Log odds and it works very well for me
This is an example
X yes no total Prob Odds Log Odds
0 4 1 5 0.8 4 1.39
1 5 2 7 0.71 2.5 0.92
2 3 2 5 0.6 1.5 0.41
3 1 2 3 0.33 0.5 -0.69
y1weights x1weights 1slope intercept
0.66 1.3 -0.65 1.50
The intercept is calculated as y1weightsts1-x1weights*1slope
the prediction follows the normal linear regression
intercept+1slope*x
I have had a idea (maybe slighty crazy) and perform some multiple regression
here is the 2nd set
X yes no total Prob Odds Log Odds
0 2 6 8 0.25 0.33 -1.10
1 3 9 12 0.25 0.33 -1.10
2 2 11 13 0.15 0.18 -1.70
3 1 15 16 0.06 0.07 -2.71
y2weights x2weights 2slope
-1.784954575 1.755102041 -0.595625389
If this was normal multiple regression it would fall like this
intercept+1slope*x1+2slope*x2
But how do I calculate the intercept so it takes in the 2nd set that I am looking at
I had thought
(y1weights-x1weights*1slope)-(y2weights-x2weights*2slope)
But I am not sure of if this is right?
Hi Again
Basically I have 2 sets of data which have the same Independent variable (X) but different dependendent variables (Y)
I would like to combine the 2 sets to create one prediction model
I have created seperate slopes and intercepts for each of the sets but I need to know how to calculate an overall intercept so I can use the normal regression formula
intercept+1slope*x1+2slope*x2
Just merge the data and do a regression, and if they are disjoint sets of data, then consider either setting the values to zero, using some averaging technique, or using existing regression models to create fitted data for the other data set and merge the two data sets together and compare the results.
But I should warn you that if both sets of data don't represent the same model, then the output you will get will largely be garbage and be useless (regression model that is).