When your Ct values are not consistent and you find such a high Ct in your data, I suggest you repeating the qPCR. Do you see that in only one of your samples or is it more frequent? Also, do you use SYBR-green or TaqMan?

If this difference between your triplicates you are experiencing is only in one of your samples, I would say you can exclude it as you say it due to technical error and that is the idea of having triplicate on your samples.

If there is some recognizeable problem in the amplification curve, then you may (you should!) exclude this value. If the amplification products are wrong, you should also exclude it.

If the correct product is amplified specifically and the amplification curve looks ok, there is not much evidence that 31.1 is the "outlier" (outlying from what? from just two other values that happen to be quite close together? that'd be a strange way of reasoning!). Sure, it may be that you accidentally lost some sample (actually > 95% of the sample!) when pipetting it into the PCR. Nobody can say. How plausible do you find this?

If the technical replication is done by preparing a mix containg the sample and then distributing it (mastermix+sample) into three PCR wells, a pipetting error leading to this difference must be visible in the volumes in the wells (you can check it).

Noticing this "outlying" value is a good indication to crosscheck the curves and the wells. If there is no reason or hint other than the fact that the value 31.1 is (quite) a bit different to 26.0, then there is no reason to exclude it. If you don't trust it, then don't trust the other two values either and re-do the PCR, as Vahid suggested.

However, I don't subscribe to Vahid's statement that we do triplicates to identify outliers. We do them to get better estimated of the Ct values, because the measurements are relatively noisy (0.2 Ct difference means already more than 10% error). Mean Ct values have a lower variance than single Ct values. This is the reason. No doubts: suspicios deviations may give us a hint that something could have gone wrong, this is something that triplicates can and should be used for. Actually, one should carefully inspect the amplification curves before calculating the Ct values, but this is not always feasable. Strange Ct values allow us to go back to the specific curves that might be problematic and check them.

I've used Dixon's Q test in the past. This is useful for finding outliers in small numbers of observations. You can get an outline of the method at https://en.wikipedia.org/wiki/Dixon's_Q_test.

Basically you put your data in numerical order, and calculate the range of the data (largest - smallest). You then calculate the gap between the presumed outlier and the next number - this is the gap.

You then calculate do the calculation: Q= gap/range.

You then look up the table below and see if the outlier is significant at the 90, 95, or 99% confidence. You just need to look at the correct column, reflecting the number of observations made.

Number of

values: 3 4 5 6 7 8 9 10

Q90%: 0.941 0.765 0.642 0.560 0.507 0.468 0.437 0.412

Q95%: 0.970 0.829 0.710 0.625 0.568 0.526 0.493 0.466

Q99%: 0.994 0.926 0.821 0.740 0.680 0.634 0.598 0.568

In your example (26.0, 26.02 and 31.1) you have a range of 31.1 - 26.0 = 5.1

The gap is 31.1 - 26.02 = 5.08.

Q = 5.08/5.1 = 0.9961 which is greater that Q99% for 3 observations. Therefore 31.1 is an outlier with 99% probability.