Hi hatem

It seems that you need something like my Multi-Frame Indicator Kit. "Multi-Frame I RSI" can return a 14-period 30-minute RSI on any chart or data from tick to 30 minutes, "Multi-Frame X RSI" can return a 14-period daily RSI on any chart or data from tick to daily, and "Multi-Frame D RSI" can return a 14-period, weekly RSI on any chart or data from tick to weekly. However, you need to understand that plotting a 14-period weekly RSI on a 15-minute chart would need 8000 to 10,000 15-minute bars to plot accurately. If it's 5-minute data that you're using you would need up to 25,000 bars to get an accurate 14-week RSI. Whichever way you look at it a 14-week RSI needs more than a year of data for any sort of accuracy. My rule-of-thumb is to multiply the Periods parameter by 5 for any function or indicator that uses Wilders or an EMA to calculate its result.

Roy

www.metastocktips.co.nz

Hi Mark

My 25000 was a rough guess but lets be very clear - you need many morbars than the minimum you've calculated because of the way an exponential moving average is constructed. RSI includes two Wilders Smoothing functions in its makeup, and Wilders(C,14) is the same as Mov(C,27,E). An EMA uses at least a fragment of ALL past data to calculate a result, so when the data available is limited it tends to come up with an answer that's somewhat different from what you'd expect.

Try plotting a weekly chart with a couple of years worth of data and read off the last value of RSI(C,14). Then reduce the loaded data to just 6 months (26 weeks) and read off the RSI value again. I just did that exercise on a chart and got values of 41.897 and 39.877. Using 70 weekly bars I got 42.106 and for 100 bars I got 41.897. OK, so when a slower Wilders is involved you should increase the minimum amount of data by a factor of 7 or more rather than 5? Multiplying the Periods parameter by a factor of 5 will give you a result that is usually going to be near enough. Where people usually get caught out is by using Minimum Periods for an exploration and wondering why the result differs significantly from what they see on their charts.

An EMA is recalculated on each bar by adding in X% of the value being smoothed (discarding X% of the old value is another way of putting it) and retaining (100-X)% of the previous EMA value. Hope this helps.

Roy