STACK DocumentationDocumentation home | Category index | Site map
An answer test is used to compare two expressions to establish whether they satisfy some mathematical criteria. The prototype test is to establish if they are the same. That is to say, algebraically equivalent.
The exact behaviour of each answer test can be seen from STACK's test suite for STACK Answer tests.
This compares pairs of expressions and displays the outcomes
from each test. Mainly used to ensure STACK is working, it is
invaluable for understanding what each test really does. In
particular it enables authors to see examples of which
expressions are the same and different together with examples
of the automatically generated feedback. This feedback can be
suppressed using the
quiet tick-box in the potential response
Informally, the answer tests have the following syntax
[Errors, Result, FeedBack, Note] = AnswerTest(StudentAnswer, TeacherAnswer, Opt)
|StudentAnswer||A CAS expression, assumed to be the student's answer.|
|TeacherAnswer||A CAS expression, assumed to be the model answer.|
|Opt||Any options which the specific answer test provides. For example, a variable, the accuracy of the numerical comparison, number of significant figures.|
Note that since the tests can provide feedback, tests which appear to be symmetrical, e.g. Algebraic Equivalence, really need to assume which expression belongs to the student and which to the teacher.
|Errors||Hopefully this will be empty!|
|FeedBack||This is a text string which is displayed to the student. It is CASText which may depend on properties of the student's answer.|
|Note||This is a text string which is used for Reporting. Each answer note is concatenated with the previous notes and any contributions from the branch.|
The feedback is only shown to a student if the quiet option is set to 'no'. If feedback is shown, then examples are given in the answer-test test suite. Login as the admin user then navigate to
Home > Site administration > Plugins > Question types > Stack
A crucial component in the assessment process the ability to decide if two expressions are equal.
Let us assume a teacher has asked a student to expand out and the response they have from one student is . This is "correct" in the sense that it is algebraically equivalent to and is in expanded form (actually two separate mathematical properties) but "incorrect" in the sense that the student has not gathered like terms by performing an addition . What about a response ? This is, arguably, better in the sense that the terms are gathered, but the student here did not order the terms to write their expression in canonical form. Hence, we need quite a number of different answer tests to establish equality in various senses of the word.
|CasEqual||Are the parse trees of the two expressions equal?|
|EqualComAss||Are they equal up to commutativity and associativity of addition and multiplication, together with their inverses minus and division? For example but This is very useful in elementary algebra, where we want the form of the answer exactly. Simplification is automatically switched off when this test is applied, otherwise it makes no sense.|
|AlgEquiv||Are they algebraically equivalent, i.e. does the difference simplify to zero?|
|SubstEquiv||Can we find a substitution of the variables of into which renders algebraically equivalent to ? If you are only interested in ignoring case sensitivity, you can apply the Maxima commands defined by STACK
|SameType||Are the two expressions of the same types_of_object? Note that this test works recursively over the entire expression.|
|SysEquiv||Do two systems of polynomial equations have the same solutions? This test determines whether two systems of multivariate polynomials, i.e. polynomials with a number of variables, generate the same ideal, equivalent to checking they have the same solutions.|
This is the most commonly used test. The pseudo code
If simplify(ex1-ex2) = 0 then true else false.
This test will work with a variety of types of object of mathematical objects, including lists, sets, equations, inequalities and matrices.
- This test disregards whether simplification is switched on, it always fully simplifies all its arguments.
AlgEquiv(predicate(ex),true)with predicate functions.
Note: exactly what it does depends on what objects are given to it. In particular the pseudo code above only applies to expressions. We cannot subtract one list or set from another, so we have to use other tests.
For sets, the CAS tries to write the expression in a canonical form. It then compares the string representations these forms to remove duplicate elements and compare sets. This is subtly different from trying to simplify the difference of two expressions to zero. For example, imagine we have and . One canonical form is to expand out both sides. While this work in principal, in practice this is much too slow for assessment.
Currently, and are considered to be different. If you want these to be considered the same, you need to write them in a canonical form. Instead of passing in just the sets, use the answer test to compare the following.
Why doesn't the test automatically apply
radcan? If we always did this, then and would be expanded out, which would break the system. Since, in a given situation, we know a lot about what a student is likely to answer we can apply an appropriate form. There isn't one rule which will work here, unfortunately.
There are also some cases which Maxima can't establish as being equivalent. For example As Maxima code
This is Cardano's example from Ars Magna, but currently the AlgEquiv test cannot establish these are equivalent. There are some other examples in the test suite which fail for mathematical reasons. In cases like this, where you know you have a number, you may need to supplement the AlgEquiv test with another numerical test.
EqualComAss: Equality up to Associativity and Commutativity
This test seeks to establish whether two expressions are the same when the basic operations of arithmetic addition/multiplication and Boolean and/or are assumed to be nouns but are commutative and associative. Hence, but . The unary minus commutes with multiplication in a way natural to establishing the required form of equivalence.
Notice that this test does not include laws of indices, so . Since we are dealing only with nouns does not simplify to . E.g. . This also means that is not considered to be equivalent to under this test. In many situations this notation is taken mean the same thing, but internally in Maxima they are represented by different functions and not converted to a canonical form by the test. Extra re-write rules could be added to achieve this, which would change the equivalence classes.
This is a particularly useful test for checking that an answer is written in a particular form, e.g. "simplified".
This test can also be used to establish , but since the arguments of the set constructor function are commutative. Sets are not associative, so . (See Maxima's
The CAS returns the result of the simple Maxima command
if StudentAnswer=TeacherAnswer then true else false.
There is no explicit simplification here (unlike AlgEquiv).
This test always assumes simplification is off, i.e.
simp:false, regardless of any question settings. If this is too strict, use
ev(ex,simp) in the arguments to simplify them explicitly first.
When simplification is off this test effectively tests whether the parse trees are identical.
Please note, the behaviour of this test relies on the internal representation of expressions by Maxima, rather than an explicit mathematical property such as "equivalence". Explicit properties should be tested in preference to using this test!
The SysEquiv (system equivalence) test takes in two lists of polynomial equations in any number of variables and determines whether the two systems have the same set of solutions. This is done using the theory of Grobner bases to determine whether the ideals generated by the two systems are equal. As the test allows for polynomials in several variables, it can cope with the intersections of the conic sections, as well as a large number of geometrically interesting curves.
- This test disregards whether simplification is switched on, it only simplifies its arguments where required. This allows the test to list equations in feedback that the student has erroneously included in their system.
This test deals with equality of sets. The algebraic equivalence functions give very minimal feedback. This test is designed to give much more detailed feedback on what is and is not included in the student's answer. Hence, this essentially tells the student what is missing. This is kind of feedback is tedious to generate without this test.
The test simplifies both sets, and does a comparison based on the simplified versions. The comparison relies on
ev(..., simp, nouns) to undertake the simplification. If you need stronger simplification (e.g. trig) then you will need to add this to the arguments of the function first.
Equiv and EquivFirstLast
These answer tests are used with equivalence reasoning. See the separate documentation.
Often, we wish to establish if the student's expression has the correct form. For example, consider the following various written forms of .
Each of these might be considered to be factored. Establishing
ex is factored is not the same as comparing it with
Related tests establish that an expression is expanded or in partial fraction form.
|LowestTerms||This test checks that all numbers written in the first expression are in lowest terms and that the denominator is clear of surds and complex numbers. Notes|
|* if you want to check whether a rational polynomial is written in lowest terms, this is not the test to use. Instead, apply the predicate functions
|* the second argument to this function is ignored, i.e. this test does not confirm algebraic equivalence. You might as well use 0 here.|
|SingleFrac||This test checks that SAns written as a "single fraction". See below.|
|Expanded||Confirms SAns is equal to
|FacForm||This test checks (i) that SAns is algebraically equivalent to TAns , and (ii) that SAns is "factored" over the rational numbers. See below for more details. The answer test expects the option to be the variable, which is needed to generate feedback. If the answer is incorrect, quite detailed feedback is provided.|
|PartFrac||This test checks (i) that SAns is algebraically equivalent to TAns , and (ii) that SAns is in "partial fraction form". The option must be the variable.|
|CompletedSquare||This test checks (i) that SAns is algebraically equivalent to TAns , and (ii) that SAns is in "completed square form". The option must be the variable.|
This test checks (i) that SAns is algebraically equivalent to TAns , and (ii) that SAns is written as a single fraction. Notes
- This test works at the top level, making sure the expression as a whole is a single fraction.
- if you also want this expression written in lowest terms, then this is quite a separate test. You need to first confirm you have a single fraction then add a new potential response. One way is to use the ../CAS/Predicate functions
lowesttermsp(ex)and compare the result with
truewith the AlgEquiv test.
- The algebraic equivalence check is for convenience. If you only want to check an expression is a single fraction make , i.e. ATSingleFrac(ex,ex) will do.
Factorisation of polynomials
An expression is said to be factored if it is written as a product of powers of distinct irreducible factors. Strictly speaking, in establishing that an expression is in factored form, we might not even care whether the terms in the product are fully simplified, as long as they are irreducible.
Irreducibility on the other hand means we can't find further factors, but here we need some care.
- Any non-trivial factorization, e.g. .
- A factorization into irreducible factors over the integers/rational numbers, i.e. .
- A factorization into terms irreducible over the reals, i.e. .
- A factorization into irreducible polynomials over the Gaussian integers, with allowed, i.e. .
- A factorization over the complex numbers, where the factor would also be split into the four terms .
In elementary teaching, meaning 4. is unlikely to occur. Indeed, we might take this example to represent factoring over any extension field of the rational numbers. We normally seek to establish that the factors are irreducible over the integers (which is equivalent to irreducibility over the rational numbers) or the reals. But, unlike a canonical form, we are not particularly interested in the order of the terms in this product, or the order of summands inside these terms.
The FacForm test establishes that the expression is factored over the rational numbers. If the coefficients of the polynomial are all real, at worst you will have quadratic irreducible terms. There are some delicate cases such as: vs and vs , which this test will cope with.
Factorisation of integers
If you would like to ask a student to factor a polynomial, then do not use the FacForm answer test. The FacForm answer test is designed to use with polynomials.
Instead, switch off simplification and define
and use EqualComAss as the answer test.
Note however that EqualComAss does not think that
2*2*3 are the same!
These tests deal with the precision of numbers. See dedicated page on numerical answer tests.
This test is a general differentiation test: it is passed if the arguments are algebraically equivalent, but gives feedback if it looks like the student has integrated instead of differentiated. The first argument is the student's answer. The second argument is the model answer. The answer test option must be the variable with respect to which differentiation is assumed to take place.
This test is designed for a general indefinite integration question: it is passed if both the arguments are indefinite integrals of the same expression. The first argument is the student's answer. The second argument is the model answer. The answer test option needs to be the variable with respect to which integration is assumed to take place, or a list (see below).
Getting this test to work in a general setting is a very difficult challenge. In particular, the test assumes that the constant of integration is expressed in a form similar to +c, although which variable used is not important.
The issue of vs is a particular challenge. The test is currently defined in such a way that if the teacher uses in their answer, then they would expect the student to do so. If they don't use the absolute value function, then they don't expect students to but will accept this in an answer. It is, after all, not "wrong". However, in the case of partial fractions where there are more than one term of the form then we insist the student is at least consistent. If the teacher has any then the student must use in all of them. If the teacher has no (i.e. just things like ) then the student must have all or none.
The Int test has various additional options.
The question author must supply these options in the form of a list
[var, opt1, ...]. The first argument of this list must be the variable with respect to which integration is taking place.
If one of the
opt? is exactly the token
NOCONST then the test will condone a lack of constant of integration. That is, if a student has missed off a constant of integration, or the answers differ by a numerical constant, then full marks will be awarded. Weird constants (e.g. ) will still be flagged up.
The answer test architecture only passes in the answer to the test. The question is not available at that point; however, the answer test has to infer exactly which expression, including the algebraic form, the teacher has set in the question. This includes stripping off constants of integration and constants of integration may occur in a number of ways, e.g. in logarithms. In many cases simply differentiating the teacher's answer is fine, in which case the question author need not worry. Where this does not work, the question author will need to supply the expression from the question in the right form as an option to the answer test. This is done simply by adding it to the list of options.
The test cannot cope with some situations. Please contact the developers when you find some of these. This test is already rather overloaded, so please don't expect every request to be accommodated!
The following tests do not use Maxima, but instead rely on PHP.
|String||This is a string match, ignoring leading and trailing white space which are stripped from all answers, using PHP's trim() function.|
|StringSloppy||This function first converts both inputs to lower case, then removes all white space from the string and finally performs a strict string comparison.|
|RegExp||A regular expression match, with the expression passed via the option. This regular expression match is performed with PHP's
|NOTE: we plan to remove this test in STACK version 4.3. Do not use this test.|
A dedicated answer test for scientific units is described on the units page.
Adding answer tests is possible, but is a developer task.
Documentation home | Category index | Site map