anonymous function + IF + fitnlm = possible?

Hello. Is it possible to have an anonymous function combined with an If statement in a non-regression function? something like this --->

    tbl = table(x1, x2, x3, y);

if table.x1 < 30
  modelfun1 = @(b,x)(...Equation (1)...);
  b0 = [1 1 1 ...];

else % table.x1 > 30
  modelfun2 = @(b,x)(...Equation (2)...);
  b0 = [1 1 1...];

    mdl1 = fitnlm(tbl, modelfun1, b0)
    mdl2 = fitnlm(tbl, modelfun2, b0)
  end

Answer 1

[Star Strider]

Not an if statement. However you can use a form of ‘logical indexing’ to do essentially the same thing.

Example —

x = linspace(0, 60);
b = [3 11];
modelfun = @(b,x) (x<=30).*(b(1).*x/10 - b(2)) + (x>30).*(b(1).*sin(x.*b(2)));

figure(1)
plot(x, modelfun(b,x))
grid

Logical vectors evaluate as (0,1) numeric values in calculations.

[wesleynotwise]

IT WORKS LIKE MAGIC in the fitnlm function too. Thanks once again.

Sidetrack 1: I assume this also applies to categorical variable? eg: (x =='apple').*(b(1)...)

Sidetrack 2: why IF statement can't be used in this case?

[Star Strider]

As always, my pleasure.

1. I never tried it with categorical variables. In any event, you will need to use the strcmp (link) or strcmpi (link) functions to do string comparisons.
2. The if block is not an option for anonymous functions due to the restrictions on anonymous function syntax. You could use if blocks in function files, however, since they use the syntax rules for scripts. The documentation on Anonymous Functions (link) offers further clarification.

[wesleynotwise]

Noted. Cheers.

[Star Strider]

Cheers.