I have 3 parameters of a machine ( independent variables) those impact on 4th one , DV ( continuous variable ). I have performed experiments as following:
Holding IV2 , IV3 constant and increasing IV1 in equal intervals and noted DV
Holding IV1, IV3 constant and increasing IV2 in equal intervals and noted DV.
Holding IV1, IV2 constant and increasing IV3 in equal intervals and noted DV.
Making combinations of IV1, IV2 and IV3 by holding any one constant and varying in equal steps rest two and noted DV
Finally, Making incremental changes in all three IVs with equal intervals and noted DV.
Aim is to find optimized value of DV with different combinations of THREE IVs along with finding most and least contributor in predicting DV.
I do not know how to go about this!
Thanks in advance
Regards
Rafique
If you really have no idea then I suggest to cooperate with a local statistician.
To help you, one would need to now of what kind your DV is, i.e. does it have boundaries?, are the values close to boundaries?, are these values like counts or proportions or concentrations?, how does the distribution look like?, etc etc.
Further it would be helpful to know more about your IVs. Are they nominal, ordinal, interval-scaled? If not interval: how many levels are there? If many: can they be treated as "random effects"? It they are metric: can the relationship between the IVs and the DV be non-linear?
Then: do you expect or do you like to consider potential interactions between the variables?
This all has to be known (investigated) in order to build a sensible model that will reasonably well describe the functional realtionship of the IVs and the DV as well as the structure of the variability (the stochastic part of the model). Such a model can then be fit to your data. You can then perform an analysis of deviance (or of variance - as a special case) to find those IVs with the largest effect on "explaining the variation in the data". The estimated coefficients give you the effect sizes (their meaning will depend on the model you will use). You can use the fitted model to deterime the combinations of IVs at which the DV has its expected maximum (or minimum) etc.
Hi Jochen,
thank you very much for detailed reply.
My IV1 has 5 values ( 1000,2000,3000,4000,5000) , IV2 has 5 values ( 100,200,300,400,500) and IV3 has 2 values ( 2, 4) . I want to predict minimum value of my DV with these combinations.
Regards
Rafique
It seems that your IVs are numeric, so you might consider some form of multiple regression. Finding an appropriate model is something that requires a thorough investigation of your data. We won't solve this through RG. So either you are fit enough to do this or you should sit together with someone who is firm in modelling (like a statistician).
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA