分割图重复测量方差分析，线性回归和R中的不同结果

Question

This question here came after a long research about the subject. I'll appreciate your reply.

I conducted a research and now my goal is to test the effect of the Treatment on the ouctome , but I want to check the moderation effect of the Race .

n treatment: 83 
n placebo: 76
No missings.

My original dataset is in Wide format.

Where:

**ID -> Participant identification
Con_dummy -> Treatment or Control (Between subjects)
Hispanic -> Yes or no (Between subjects)
M41 and M_42 -> Dependent variables. M_41 was measured in 2016 and M_42 was measures in 2017.**

I transformed my dataset to a long format using this code:

 d2_stack_long <- df_stack %>% group_by(cond_dummy) %>% select(ID, cond_dummy, hispanic, M_4I1TOT,M_4I2TOT) %>% gather(key="ID", value, M_4I1TOT,M_4I2TOT) %>% tbl_df() %>% setNames(c("ID","Condition","Hispanic", "Time","Result")) %>% mutate(Time = ifelse(Time == "M_4I1TOT", "Time 1", "Time 2")) %>% arrange(ID) 

I've got a message

Warning message: attributes are not identical across measure variables; they will be dropped

But now I have a long dataset. I have the double numbers of rows now, because every participant was measured at two time point, but everything remains the same

 > d2_stack_long %>% group_by(Condition, Time) %>% summarise(mean(Result),n=n()) # A tibble: 4 x 4 # Groups: Condition [?] Condition Time `mean(Result)` n <chr> <chr> <dbl> <int> 1 Control Time 1 5.973684 76 2 Control Time 2 6.342105 76 3 Treatment Time 1 6.277108 83 4 Treatment Time 2 9.626506 83 

When, now, I test this model, I achieve two different results when using R and when using JASP.

mod <- lm(data=d2_stack_long , Result ~ Time + Condition*Hispanic)
Anova(mod, type=3)

I think I'm forgeting to adjust something and I'll appreciate your help.

I read several posts before asking this question, but I just found people commenting about math equation to ANOVA or comparing results across packages like here .
A famous blog with examples is down.

Thanks.

Answer 1

If someone wants the answer. After a Long journey I got from this post . The sums of squares type in R is set to 1, whereas in JASP it is set to 3.

You can add an ERROR term in Regression models (Repeated measures), despite 99% of posts here.

options(contrasts=c("contr.helmert", "contr.poly"))
anova(lm(Result ~ Time*Condition*Hispanic*ID, data=d2_pals_total))

And if you want to go with aov funcion:

summary(aov(data=d2_pals_total, Result ~ Time*Condition*Hispanic +
      Error(ID)))

Thanks