Independent Robust T-Test

#1
Hi all,

I'm attempting to compare differences between 2 unequal group sizes (one ~ 97 the other ~ 714). The reason for the large discrepancy is I am looking at a program done by one class to see if it is significantly different than what has occurred in previous classes. I've been reading about robust stats recently and decided to use a yuen bootstrap in R-Studio from the WRS2 package for a more valid comparison, especially with the difference in sample size.

My formula is yuenbt(DataExample$PT500 ~ DataExample3$ClassPT500, tr = 0.2, nboot = 599, side = TRUE)

and it returns

Call:
yuenbt(formula = DataExample$PT500 ~ DataExample$ClassPT500,
tr = 0.2, nboot = 599, side = TRUE)

Test statistic: NA (df = NA), p-value = 0

Trimmed mean difference: -65
95 percent confidence interval:
NA NA

The NA's return on other variables that I've tried out as well, or in some cases the confidence interval will state INF. Any ideas why this is happening (such a big difference in sample size?) and suggestions on what the next best step would be are greatly appreciated. I've updated a spreadsheet with some of the variables.

Thank you in advance.
 

Attachments

hlsmith

Less is more. Stay pure. Stay poor.
#2
Your data doesn't make sense to me given the model. First, I don't see the variables you used in the dataset, second, I don't see any binary categorical variable, third, I see some missingness. I am guessing Pre and Post was your categorical variable, but are these data setup correctly to run the function? INF likely represents infinity. Is the effect size huge in this scenario given the dispersion metric?
 
#3
Your data doesn't make sense to me given the model. First, I don't see the variables you used in the dataset, second, I don't see any binary categorical variable, third, I see some missingness. I am guessing Pre and Post was your categorical variable, but are these data setup correctly to run the function? INF likely represents infinity. Is the effect size huge in this scenario given the dispersion metric?
Sorry for the example above it would be DataExample$PT and DataExample$ClassPT.

The Pre and Post measures are results before and after an academy. The PT variables are the differences between the two. Some people did not make it through the academy (left, were kicked out etc) so the difference and Post measures may have missing data. To my understanding these will omit NA values and it is in the correct format.

I don't believe the effect size is that large. When just comparing within group t-test, it did appear that the "Class" group saw a larger increase, but this is why I wanted to compare it to the overall to see if the difference was significant and then again to see if the starting point between the two differed and the ending point.

Apologies for any confusion, stats it's not a strong suit and I probably know just enough to mess it up!