Mixed Models in R: Missing rows vs.empty cells

I'm fitting a mixed model where measurements are repeated daily for each participant (for several variables). The data are in a format where each participant has a row for each day of the study.

Instead of leaving cells empty for missing data empty and keeping other data for each day, the person entering the data deleted the entire row for the day if any variable was missing. They didn't delete the entire record for the participant, they just deleted the entire day.

How will this change parameter estimates compared to a dataset where days with incomplete data were kept? basically I'd like to know more about what lmer() or lme() are doing under the hood. If I have 60 full entries for day 1-3, but only 50 full entries for day 4 (10 completely missing rows), how will that differ from a model where I have 50 full entries and 10 partial entries for day 4? I've learned a little bit about Maximum Likelihood Estimation, but not enough to have any practical sense of what is going on.

• lmer and nlmer remove the whole observation (= data.frame row) if it contains missing values. Fortunately, they are not very sensitive to unbalanced data. However, that the observations have been removed during data recording prevents you from imputing missing values or using more advanced mixed model implementations that can deal with missing data (I've heard of these at this site but never used them). – Roland Nov 15 '17 at 11:35
• If I'm understanding you correctly, lmer does automatically what my colleague did by hand? Is there a search term for these more advanced mixed model implementations? If I can recover the missing data, I'd like to make use of it. – dmacfour Nov 15 '17 at 17:57
• I don't know a search term. AFAIK there is no R implementation. – Roland Nov 15 '17 at 19:05