How to Calculate the Upper Prediction Limit of Data

If all data values were the same, there would be no need to calculate an average, because you would immediately recognize the trend. However, real data is rarely uniform. Data fluctuates regularly, so an average is necessary to find the central value. Unfortunately, this fluctuation in data values makes averages unreliable as predictors of future values, because without a measurement of data variability, you can't know how much deviation from the average is normal. This is where confidence intervals or prediction limits come in. Using a variability measurement, you can construct upper and lower values within which future values are predicted to fall.

Instructions

    • 1

      Add all the values in a series of data points and divide the total by the number of data points. This calculates the average. As an example, the numeric series 50, 55, 48, 56 and 52 total 258. Dividing by 5 gives an average of 51.6.

    • 2

      Subtract the average from each data point. In the example, 50 minus 51.6 equals -1.6. Likewise, subtracting 51.6 from the other four values results in 3.4, -6.6, 4.4 and 0.4, respectively.

    • 3

      Square the previous results. In the example, -1.6 times -1.6 gives you 2.56. Likewise, squaring the remaining four values gives you 11.56, 43.56, 19.36 and 0.16.

    • 4

      Add the squared values. In the example, adding 2.56, 11.56, 43.56, 19.36 and 0.16 gives you a total of 77.1.

    • 5

      Divide this total by the number of data points, minus one. In the example, 77.2 divided by 4 gives you 19.3.

    • 6

      Take the square root of the previous calculation to calculate the standard deviation. In the example, taking the square root of 19.3 gives you a standard deviation of 4.39.

    • 7

      Divide the standard deviation by the square root of the number of data points. This calculates the standard error. In the example, the square root of 5 is 2.24. Dividing this into 4.39 gives you a standard error of 1.96.

    • 8

      Multiply the standard error by the number of standard errors you wish to use. The most common method is to use two standard errors, which creates a confidence interval within which 95 percent of all future measurements are expected to fall. In the example, this results in 3.93.

    • 9

      Add this figure to the average to calculate the upper prediction limit. Subtract it from the average to calculate the lower limit. In the example, 95 percent of future values are expected to be between the upper prediction limit of 55.53 and the lower limit of 47.67.

Learnify Hub © www.0685.com All Rights Reserved