Forum: support

RE: using cloze with exams2nops [ Reply ]
By: Achim Zeileis on 2020-02-14 21:16

Cloze questions are not supported in exams2nops() and it wouldn't be straightforward adding support for these. However, you can simply create a series of {schoice} questions, say 1a.Rnw, 1b.Rnw, 1c.Rnw, 1d.Rnw. If 1a.Rnw generates a certain vector with data, then the data can be re-used in 1b.Rnw etc.

If you use .Rmd (rather than .Rnw) exercises, then you just have to set envir=.GlobalEnv so that the random data is generated in the global environment where the subsequent exercises can find it again. (For .Rnw exercises this is the default and only option anyway.)

If you want the exam to be fully automatically evaluated, then the question 1c about the test statistics has to be phrased as a single-choice question, i.e., the correct statistic along with some false statistics (either generated randomly or via typical mistakes).

Instead of the single-choice question it is possible to include a {num} question. However, this will generate a second evaluation sheet that has to be corrected manually before it can be scanned.

using cloze with exams2nops [ Reply ]
By: jpmaroco Maroco on 2020-02-14 15:23

[forum:47396]

Dear Achim,

Is it possible (maybe with some dev version of Exams) to have cloze type items with exams2nops? (What can I do, I like the automatic correction).

Or... can I just write a common stem for a set of 4 schoice/numeric questions? For example, in a statistics block of 4 items:

1. Consider the following weight data for two groups (control and experimental groups) [A table with two columns shows the data]. Considering that the data is normally distributed and the variances are homogeneous.

1.1 Which of the following the most powerful test to find out if the differences between groups are statistical significant?
(a) t-test
(b) z-test
(c) Mann-Whitney
(d) Fisher

1.2. Which are the correct H0 for the most power full test?
(a) H0: m1=m2
(b) H0: s1=s2
(c) H0: p1=p2
(d) H0: X~N(m,s)

1.3. The value of the test statistic (round the result to 3 decimal places) is:
[_____]

1.4. For alpha=0.05 we can conclude that the differences are statistical significant
(a) True
(b) False

Any idea on how this could be implemented with exams2nops and automatic grading?

Thanks (also for the great R/Exams)!
João Marôco