why not using (your mentioned) regression modelling, e.g. a generalised linear one with multinomial response. To receive more precise answers, you may tell us more about your data.

You should use demand models based on linear regression models (such the Kenesian model of consumption) or the economic discrete choice models (probit and logit models) such as the choice of modes of transport or Random Utility Functions (McFadden(1978)).

thanks, colleagues. I have a lot of data. such as...consumption of each item in market basket.. as well I have income of households and. expenditure of the households on the food stuff and non-alcoholic drinks.. and final dailz recommended dose for each food item

An important consideration is if you have individual household data and are trying to model an individual consumer's behavior as opposed to aggregated data that would be used to model consumers' behavior. I am guessing you have the latter.

Often, the type of data we have limits what we can do.  For example, without data on price, or at least price-like data, our ability to construct demand functions is seemingly impossible.  however, that does not prevent us from using another source for prices or even using surrogates for prices.

Even without price data, observing trends over time is possible, for example. Some of this can overly simplified, while sometimes additional precision may be possible; for example, the use of ARIMA and other models can sometimes help us to estimate turning points, which are elusive at best in other models.  I hope this gives you some ideas.

Thank Rajesh, I checked youtube. It si theoretical, but for the understanding is suitable. I found very interesting pdf file programming in SAS http://support.sas.com/techsup/technote/mr2010f.pdf. I can imagine to apply my dataset to choice model.

In non contractual setting, in wich the customer defection is not observed, you can use the Pareto/NBD or the BG/NBD models.

It depends on your theoretical framework, but I think structural equation models are also a good option. 

Seeing your problem, most prevalent is the study of discrete choice (McFadden, 1974), which like the work on demand model, also emphasizes the direct and close connection between economic theory, econometrics and empirical work. See McFadden (1981, 1984) and Train (2003) for surveys of this line of research.

Regards,

Mohit Jamwal

There are many other models for modelling consumer behaviour. Most common is the Recency, Frequency and Monetary model. Models involving enterprise activity and customer data have been found to be more accurate than econometric models. The econometric models that rely on GDP, GNP, unemployment rates, etc are typically used for devising national policies. If you are interested in devising strategies for your business, I recommend you use RFM, Customer Lifetime Value Models, Customer Segmentation Models, etc

I would recommend „Handbook of Marketing Decision Models” by B. Wierenga (Ed.)

See "The Dynamics of Household Budget Allocation to Food Expenditures" by J.Benus, J. Kmenta and H.Shapiro in the Review of Economic and Statistics 58 (May 1976) pp.129-138

You could estimate Almost Ideal Demand System using seemingly unrelated regression (SUR) model.

Have a look at my thesis = Working Paper 98 at www.NIER.se

They have a working paper series at the NATIONAL INSTITUTE OF ECONOMIC RESEARCH.

ALL TE BEST BHARAT BAROT