Additional info based on the comments below:
It looks like you have a timeseries of ETF data. If you want to model it as a timeseries and do forecasting, you may find it helpful to follow this example of time series forecasting using an ensemble of boosted regression trees: https://www.mathworks.com/help/stats/time-series-forecasting-using-ensemble-of-boosted-regression-trees.html
For a broader look at options for financial modeling, you could consider other tools in the MATLAB computational finance suite (https://www.mathworks.com/solutions/computational-finance/computational-finance-suite.html) including the financial modeling and econometrics toolboxes. The computational finance suite includes tools to "perform time-series analysis and create predictive models."
For regression problems, it is true that the "Response variable must be a numeric vector". The first step for any simple regression modeling is to prepare your data, as described at https://www.mathworks.com/help/stats/linear-regression-model-workflow.html . Once you prepare the data, you can answer these questions:
- How many observations are there in the data set? (Reference the example https://www.mathworks.com/help/stats/time-series-forecasting-using-ensemble-of-boosted-regression-trees.html for one way of creating a table of observations-with-lag from a timeseries of measurements, if that is something that you want to do with your data.)
- How many predictor variables are there? What is the datatype of each predictor?
- What is your response variable? That is, what are you trying to predict? What is the datatype of the response variable? How many unique response variable values are there in the dataset?
First answer:
If your dataset has many missing values, consider using a different type of regression model which has better support for missing values. For example, regression trees built with fitrtree (see https://www.mathworks.com/help/stats/fitrtree.html) or an ensemble of trees built with fitrensemble (https://www.mathworks.com/help/stats/fitrensemble.html) will make use of observations that have some missing values in the predictors. These specific sections of the fitrtree doc mention the handling of missing values in the input predictor data X and the response data Y:
- https://www.mathworks.com/help/stats/fitrtree.html#bt6cr84_sep_shared-X
- https://www.mathworks.com/help/stats/fitrtree.html#bt6cr84_sep_shared-Y
Ordinary Least Squares (OLS) fitting can be done with the fitlm function (https://www.mathworks.com/help/stats/fitlm.html). The fitlm function will not use observations that contain missing values. So, if many or most of your observations include missing values, this is probably not the best choice of model type for your dataset. As mentioned earlier, consider using tree models. You could also consider methods of predictor imputation prior to building an OLS model.
If this answer is helpful for you, please remember to accept this answer.
