I have a very large data set (127,000 observations, each point has 22 variables). I am only interested in 7 of those variables. The two dependent variables (linear and rotational acceleration) are continuous, while the rest are categorical.
- Position = 7 factors
- Seniority = 5 factors
- Session = 2 factors
- Season = 6 factors
- Player = 64 different subjects
I would like to include a linear mixed effects analysis of the Linear and Rotational Accelerations across Position, Seniority, and Session (Game or Practice) with Player and Season as random intercepts. My code is below, however I receive the warning message
“ Model failed to converge with max|grad| = 0.0197313 (tol = 0.002, component 1)”
when I try to create either of the models. Is this message appearing because the file is too large? I appreciate that I have over 127,000 impacts in this file, with 22 different variables for each impact – not all of which are needed.
Any guidance in this analysis would be greatly appreciated, as I am still new to using lmer in R.
ThreeYearStudyImpacts$Location <- factor(ThreeYearStudyImpacts$Location) ThreeYearStudyImpacts$Season <- factor(ThreeYearStudyImpacts$Season) ThreeYearStudyImpacts$ID <- factor(ThreeYearStudyImpacts$Player.ID) ThreeYearStudyImpacts$Position <- factor(ThreeYearStudyImpacts$Position) ThreeYearStudyImpacts$Session <- factor(ThreeYearStudyImpacts$Session) ThreeYearStudyImpacts$Seniority <- factor(ThreeYearStudyImpacts$Seniority) modelLin <- lmer(gForce ~ Position + Seniority + Session + (1|ID) + (1|Season), data = ThreeYearStudyImpacts) modelRot <- lmer(Rotational.Acceleration ~ Position + Seniority + Session + (1|ID) + (1|Season), data = ThreeYearStudyImpacts)