How to run a fitlm regression for day of the week dummy variables for financial return data

Hi, I am struggling with how to set up my data to run the fitlm function for day of the week dummy variables.

My data that I have set up in a (2500,6) matrix where C1 is the returns data and C2:C6 are column vectors of zeros and ones where C2 - 1 is for Monday C3 - 1 represents tuesday and so on.

My reults are as follows:

 Fitlm:
 -1.31153554169770  0.953057779050930  -1.37613434413573  0.169037651868845
 -0.131905728387301  0.0695763306138066  -1.89584197993226  0.0582218112735167
 0.0843907252336952  0.0872037277677016  0.967742175638410  0.333368798643977
 0                  0                    NaN                   NaN
 -0.0247423867293685  0.0984537148158696  -0.251309834023453  0.801617743803277
 0.169341528221423  0.0952957370120605  1.77701053091169  0.0758206160733013
 0.460051465197619  0.0967773549107878  4.75370984897974  2.23973346792996e-06

How do i correct my input to replace this row of Nan values?

Answer 1

[Spruha]

Hello harry,

As you mentioned, you are experiencing issues with data setup for “fitml” function, resulting in the output containing NaN.

When you include dummy variables for all categories of a categorical variable (like days of the week) without a reference category, it leads to perfect multicollinearity.

To avoid this, one category is usually omitted, and its effect is captured in the intercept.

For more details about “fitml”, refer documentation:https://www.mathworks.com/help/stats/fitlm.html

Hope this helps.