Why does fit type not accept my convoluted function as input? (MATLAB)

The ultimate goal is to fit a costum equation to experimental data by (Z = f(X,Y)). The general equation is given by:

Rw = conv2([zeros(4095,11);((1/(3*redmu+mu_a))^(-3/2)).*(X).^(-5/2).*...
            exp(-mu_a*2.*(X)).*exp(-(Y.^2)./(4/(3*redmu+mu_a)*2.*X))],IRF,'valid');

Where mu_a and redmu are the fitting parameters.

The generalized piece of code I wrote can be seen below

IRF = rand(4096,1);
X = (1:4096)';
Y = 1:11;
ft_notworking = fittype(@(mu_a,redmu,X,Y) conv2([zeros(4095,11);((1/(3*redmu+mu_a))^(-3/2)).*(X).^(-5/2).*...
        exp(-mu_a*2.*(X)).*exp(-(Y.^2)./(4/(3*redmu+mu_a)*2.*X))],IRF,'valid'),'independent',{'X','Y'},'coefficient',{'mu_a','redmu'});   

It gives an error: "Dimensions of arrays being concatenatedare not consistent.". That is strange as I use the

zeros(4095,11)

and

shape = 'valid'

for the conv2 function to make sure the shape of Rw = 4096x11. Just as the input parameters `X` and `Y` are spanning. If I get rid of the convolution (as shown below) I don't get any error. This makes me suspect that the convolution and fittype function don't work together very well.

ft = fittype(@(mu_a,redmu,x,y) (((1/(3*redmu+mu_a))^(-3/2)).*x.^(-5/2).*...
        exp(-mu_a*2.*(x)).*exp(-(y.^2)./(4/(3*redmu+mu_a)*2.*x))),...
        'independent',{'x','y'},'coefficient',{'mu_a','redmu'})

What I don't get is this: Both Rw and the unconvoluted equation have the same shape `4096x11`, but for Rw we get an error and for the other we don't.

Additionally I tried to write a function that calculates Rw (the convoluted function) in the desired dimensions 4096x11 and named it test.m. I call the function in the fittype function.

ft = fittype('test( tw, rw, mu_a,redmu)','independent',{'tw','rw'},'coefficient',{'mu_a','redmu'})

I get the same error as before. So it looks like the convolution messes it up. That is unfortunate as I need the convolution. Is anyone able to help me with this problem?

Best regards

P.S Here is a script you could use to test stuff

redmu = 10;
mu_a = 0;

IRF = rand(4096,1);
X = (1:4096)';
Y = 1:11;

Rw = conv2([zeros(4095,11);((1/(3*redmu+mu_a))^(-3/2)).*(X).^(-5/2).*...
    exp(-mu_a*2.*(X)).*exp(-(Y.^2)./(4/(3*redmu+mu_a)*2.*X))],IRF,'valid');   

ft_notworking = fittype(@(mu_a,redmu,X,Y) conv2([zeros(4095,11);((1/(3*redmu+mu_a))^(-3/2)).*(X).^(-5/2).*...
    exp(-mu_a*2.*(X)).*exp(-(Y.^2)./(4/(3*redmu+mu_a)*2.*X))],IRF,'valid'),'independent',{'X','Y'},'coefficient',{'mu_a','redmu'});

ft_working = fittype(@(mu_a,redmu,X,Y) (((1/(3*redmu+mu_a))^(-3/2)).*(X).^(-5/2).*...
    exp(-mu_a*2.*(X)).*exp(-(Y.^2)./(4/(3*redmu+mu_a)*2.*X))),'independent',{'X','Y'},'coefficient',{'mu_a','redmu'});