2
$\begingroup$

I want to analysis using lmer, glmer in R.

There is variables. In fact, I have more variables like that. I think, the data cases are sufficient.

random effect : item, test, id

using lmer; random intercept : (1+item|id);

random slope : (1+test|id);

categorical independent variables(range) : sex(1,2), item(1:20), test(1:3);I used factor()

sex=factor(sex, levels=(1:2))
item=factor(item, levels=(1:20))
test=factor(test, levels=(1:3))

independent variable : age, A_score(fixed effect)

dependent variable : test score

result=lmer(score~sex+age+item+test++A_score+(1+item|id)+(1+test|id),data)

using glmer;

random intercept, random slope, independent variables are same.

dependent variable : score group(0,1)

result=glmer(score_group~sex+age+item+test+A_score+(1+item|id)+(1+test|id),data, family="binomial")

Is it correct? Also, if categorical dependent variable have 3 levels, what code can I use?

$\endgroup$
  • $\begingroup$ I think the random intercept could be fitted for id (1|id) and any variables could be considered as variability of slopes(random slope). In addition, (item|id) and (test|id) were considered as random slopes. I recommend this code: result=glmer(score_group~factor(sex)+age+item+test+(1+item+test|id),data, family="binomial") (variable sex is a cateogorical variable, andthen you should use r function 'factor'.). Your model depends on the strategy of analysis. Which variables are considered as random or fixed effect and random intercept or random slope? $\endgroup$ – J-H Yoon Jan 16 '18 at 5:25
  • $\begingroup$ yes, I use factor() for [sex, item, test] and [item, test, id] are random effect. a_score is fixed effect. $\endgroup$ – DH.LEE Jan 16 '18 at 5:43
  • $\begingroup$ First of all, I recommend to convert each categorical variables into a factor. You can use this example code:data$item_factor = as.factor(data$item). If id was considered as a random intercept and item was considered as a random slope according to the id, you might try this code:result=glmer(score_group~sex_factor+age+item_factor+(1+item_factor|id),data, family="binomial"). This code is a simple form. You can expand the code for your purpose. $\endgroup$ – J-H Yoon Jan 16 '18 at 7:11
  • $\begingroup$ I really appreciate your reply, but my professor suggested to use like that,, random intercept(1+item|id) / random slope(1+test|id). If I have to use these two things, what should I do? $\endgroup$ – DH.LEE Jan 16 '18 at 12:35