Sorry for mistake, i have two groups: treated and not, the treated group consist in children intollerant to lactos and the control are not intollerant. So i would like to investigate if into the treat group the gravity of intollerance, bmi,age and sex influence the speed of growth but i have only 11 patients.
It is called multiple regressionn. It does seem not make sense to perform multiple regression
with 4 predictors and only n=11 observations.
Generally speaking, what do you expect from complex statistical analyes based on n=11
in fact I didn't think it made sense but I wasn't sure. however, do you think there would be another type of analysis to be performed on only 11 patients to understand if the severity of the disease can have an effect on the speed of growth?
You probably don't have enough data to power a simple linear regression and meet its assumptions, and definitely not enough to control for patient characteristics in the model. A general rule which may not always hold, is you need 10 or more outcome observations for each predictor. To put it in perspective, you may have one male baby with a high BMI and certain age in the treatment group. What conclusions can you make off a single baby, none with confidence. Also, child measurements are typically not evenly collected, so if you want to say growth at 1 year, etc. do all kids really show up at exactly 1 year, if not you would need to also control for discrepancies between time intervals between growth measurement.
A doctor suggests to me to perform an ANCOVA with multiple covariates...but i don't understand 1) the differences between ANCOVA and TWO WAY ANOVA with interaction and 2) differences between ANCOVA and PLS regression. Sorry but it is the first time for me with these models
I assume ANCOVA means you have multiple predictors not just one. I don't know what PLS regression is.
With 11 cases you have multiple problems. The power of your test will be small. You may not be able to generalize to a larger population (a separate issue than power). Violations of regression assumptions will be more serve with so few cases.