One needs to apply the linear regression function to each asset separately. The data (close prices in this case) are passed to the custom factor's compute method as a 2D numpy array. There is a row for each date and a column for each asset. One needs to apply the linear regression function to each column to get a slope for each asset. The best way to apply a function to each asset is to use the numpy apply_along_axis method (https://docs.scipy.org/doc/numpy/reference/generated/numpy.apply_along_axis.html).

So your (slightly modified) code works with the addition of this method in the last line

class SlopeAdj(CustomFactor):  
    # Default inputs  
    inputs = [USEquityPricing.close]  
    window_length = 100

    # Compute SLope Adj  
    def compute(self, today, assets, out, close):  
        window_length = 100  
        def linreg(y):  
            # define X - just index on X axis  
            x = len(y)  
            X = range(1, (x + 1), 1)

            X = sm.add_constant(X) # aparently needed. adds column of ones to array  
            # define Y  
            Y = y.tolist()  
            flat_list_prices = []

            for item in Y:  
                flat_list_prices.append(item) # don't need the [0]  
            Y = [math.log(i) for i in flat_list_prices]  
            # do lin reg  
            model = regression.linear_model.OLS(Y, X).fit()  
            slope = model.params[1] # this is how to access b slope parameter in the model object  
            return slope

        out[:] = np.apply_along_axis(linreg, 0, close)

However, this can be simplified a bit. Since the inputs are numpy arrays it's fast and easy to use numpy methods rather than straight python. Both creating the X array and then getting the log prices can each be done efficiently with numpy methods. Here's an equivalent custom factor using numpy

class Slope_Using_Numpy(CustomFactor):  
    # Default inputs  
    inputs = [USEquityPricing.close]  
    window_length = 100  
    def compute(self, today, asset_ids, out, close_prices):  
        # Define the function one would like to apply for each asset  
        # The function will be passed a 1D numpy array of values  
        # In this case the close prices for past 100 days  
        def slope_of_log_prices(prices):  
            # create an array to regress against having a length equal to the price data length  
            X = np.arange(self.window_length)  
            X = sm.add_constant(X) # Adds column of ones to array for intercept  
            # Want to get the slope of the log prices so find the log values  
            log_prices = np.log(prices)

            # do linear regression using statsmodels OLS  
            model = regression.linear_model.OLS(log_prices, X).fit()  
            slope = model.params[1] # this is how to access b slope parameter in the model object  
            return slope  


        # Apply the above function across each column and output the values  
        # The zero means axis 0 which passes columns of the ndarray 'close_prices'  
        out[:] = np.apply_along_axis(slope_of_log_prices, 0, close_prices)

See the attached notebook. You may also want to check out this post for more info on the apply_along_axis method. https://www.quantopian.com/posts/custom-factor-1

Thanks a lot Dan. It works and is indeed very helpful. Thanks for the links as well