See if this works for you.

highs_15m = data.history(context.spylong,'high',390,'1m').resample('15T', label='right', closed='right').last()  
lows_15m = data.history(context.spylong,'low',390,'1m').resample('15T', label='right', closed='right').last()  


highlow = 14  
rocperiod = 12  


ema_high = talib.EMA(highs_15m.dropna(), timeperiod=highlow)  
ema_low = talib.EMA(lows_15m.dropna(), timeperiod=highlow)  
ema_mid = (ema_high + ema_low) / 2.0  
roc18 = talib.ROC(ema_mid, timeperiod=rocperiod)

A few things. You don't need to iterate with a for loop to get the average of the two ema series. You can manipulate numpy series directly. Also, the 'rocperiod' needs to be 12 or less. Otherwise you will need to fetch more data. Remember that 'roc18' is a numpy series. Not sure what you want to do with it but you may want just the last value of the series (ie roc18[-1]). That will be a scaler value.

Good luck.

Thank you for that, Dan; that was exactly what I needed although I'm curious to ask:

say, I wanted to feed the talib.ROC a list of scalar values, e.g. [1,2,4,8,16,32,64], to calculate the period=3 RoC.
Is there a way I can convert a regular list of scalar values into a pandas dataframe or whatever it needs to be to fit into talib.ROC?

The talib functions expect a numpy array as their inputs. Moreover the values of the array should be real (not integer numbers). A numpy array can be made many ways but a simple one is the 'array' method (see https://docs.scipy.org/doc/numpy/reference/generated/numpy.array.html)

my_np_array = np.array([1.0,2.0,4.0,8.0,16.0,32.0,64.0])  
roc18 = talib.ROC(my_np_array, timeperiod=rocperiod)

Notice the decimal values in the array. This will ensure it's a real array. Also, ensure 'rocperiod' is less than or equal to the length of the input array.

Hope that helps.

Ah, I see. I attempted that but one of the answers on stackoverflow told me to use np.asarray() instead of np.array().
Thank you again Dan for your miracles.