....i need some insight please...

those familiar with conjoint, though i don't have any statistical package to run on it, so my plan was to do it on an excel sheet.

Experiment set up: 6 attributes, with 3 levels each.

18 devised combinations to be ranked by respondents.

after a single respondent ranks them, i can run a multiple regression, with the independent variables as the attribut's levels (i.e 18-6=12) and the dependent as the ranks given by the respondent for each of the 18 combinations of this model. Thus the coefficients will represent the part-worths with respect to the deleted level.

my question is what next can i do, if i have 75 respondents ranked the 18 combinantions, to what extent can i work on this without a stat package!

I may be wrong .But if there are 6 attributes with 3 levels each then there are 3*3*3*3*3*3=729 possible combinations.In conjoint analysis one has to do multiple regression by ranking all 729 combvinations. Define values for each attribute.Regress ranks on values.I do not know how to use EXCEL for that.

Concerning what previous post says, I want to say that for conjoint analysis, you can not do regression all the 700-odd combinations; in stead, find out orthogonal data set in stead of so many; if the original poster has 18 after orthogonal analysis, that may be a good approach. ( I don't know much about Conjoint analysis.)

Dr you are right. For all 18 combinations 1 can regress & find part utilities.

With no stats package you can do it with Excel/Open Office (though there are also open source stats package). The way you have designed the conjoint analysis, you would analyse each individual separately (ie 75 separate regressions - if you're very good with match and offset you can do this very quickly in Excel), then each respondent has their own set of part worths that you can aggregate to an overall average/market model level.