Forgive me if this is a bad question, but I'm a newbie who get's confused with terminology frequently.

I am seeking to plan an experiment based on a few factors:

-continuous response variable (heart rate) -treatment (a drug - 3 levels) -sleep (survey) -mental state (survey)

Two Questions:

  1. What type of model should I be focusing on? Since I'm not controlling the factors from the surveys does that make this a multivariate design? If I choose to categorize the survey data into levels, is that what makes the model a simple multiple regression?

  2. Should I be adding some other factors such as gender or sex? I noticed that is common practice. Does that have to do with block design?

  • 1
    $\begingroup$ No forgiveness needed. I've been at this for a while now and I still find terminology that confuses me all the time. $\endgroup$ Aug 2 '16 at 20:30
  1. Type of model: since you are including variables from a survey (given out using a sampling technique - simple random sampling, stratified sampling, cluster, etc) you may want to use methods for survey-weighted regression (See Complex Surveys, A guide to analysis using R). It would be a linear model if the response variable is continuous; logistic if dichotomous/binary; multinomial/ordinal if there are multiple levels of the response variable. Multiple regression is simply a multivariate (more than one independent variable) regression model.
  2. Most medical studies include demographic variables (gender, race/ethnicity, age, etc) and socioeconomic status if available. As for block design, I'm not very familiar with that terminology but it seems like it is somewhat similar to controlling for confounding by including demographic/socioeconomic variables and other variables we not be interested in but could be responsible for the "trend" in the data.

I hope this helps a little

  • $\begingroup$ In psychology we consider multivariate designs to be models with more than 1 dependent variable...is the definition different for statistics/mathematics? $\endgroup$
    – Simon
    Aug 2 '16 at 22:48
  • $\begingroup$ From what i understand, gender and sex are used as blocking factors, then we have a block of subjects that are defined by the factors. This brings my next concern, this adds to the number of groups since that's one continuous variable(age) and one 2-level(sex). So in a power analysis, this requires me to have substantially more subjects. $\endgroup$
    – jmaturner
    Aug 3 '16 at 22:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.