If you report an experimentally-determined value then it should have an uncertainty (error) to go along with it. We typically report error in one of two ways: as a number or a plotted point. The former requires error to be reported explicitly (as in 23 +/- 2 mL/s) whereas the latter uses error bars to graphically indicate the error.
Whenever you report an error you should also include a brief description of how the error was determined, but you need do this only once per error term. For example, if you're reporting tank concentration measurements and you had to use a calibration curve then the first measurement you report should be something like, "The tank concentration was 100 +/- 8 ppm, where the uncertainty is the prediction interval of the equipment calibration curve." Subsequent reporting of concentrations made in the same way can be reported without the explanation.
There are four common sources of error in our lab experiments:
Measured variables. If you use a piece of equipment to directly measure a variable then you can usually find an error estimate on the equipment itself. For example, graduated cylinders usually have the words "Volume (mL) +/- 5%" printed right on the cylinder. If you're using equipment that doesn't provide uncertainty then the error is half of the smallest quantified value (e.g., if a ruler has marks every 1 mm then the error is 0.5 mm, or if a scale reads to three digits of precision (0.000) then the error is 0.0005). If a quantity is calculated from several measured variables then you'll need to perform propagation of error (see below) to report the total error. In both cases, the first such error reported would include a descriptive phrase such as, "The flow rate was 5 ± 1 mL/s, where the uncertainty results from propagation of equipment uncertainties."
Coefficients from calibration curves. If you use a piece of equipment to directly measure a variable but you have to convert the output signal of the equipment to a meaningful quantity by means of a calibration curve then you should report the error using the prediction intervals for the calibration curve. For example, if you measured the salinity of a tank by means of a conductivity probe then the probe output will have been in uS/cm (or something similar) and you'll have needed to convert to a meaningful concentration such as ppm by means of a calibration curve. In this case you'd report a measured salt concentration as something like, "The tank salt concentration was 1050 ± 20 ppm."
Repeated measurements. If you make three or more independent measurements of a variable then you can calculate a 95% confidence interval on the mean and use that as your error estimate. Confidence intervals are preferred over standard deviations because the confidence interval indicates the probability that the reported interval contains the true value of the parameter (usually a mean or regression coefficient), whereas the standard deviation only describes the distribution of the sampled data (which may or may not be a good estimate of the distribution of the true data). For example, if you repeat the synthesis of liposome nanoparticles three times under the same conditions (and therefore expect the same particle size) then you could say something like, "The mean particle diameter was 160 ± 30 nm where the uncertainty is the 95% confidence interval of the mean."
Coefficients from regression analysis. In most experiments you'll produce sets of data for which theoretical correlations exist, such as the diffusion constants in the RO experiment or the rate constants in the UV Photocatalysis experiment. In these cases the coefficients that result from a regression analysis of the data are the parameters of interest and the error should be reported as the 95% confidence interval of the regression coefficient. In cases where the slope is related but not directly equal to the parameter of interest (as is the case with some RO or UV coefficients) then you'll have to apply propagation of error rules as well to "extract" the desired value. For example, you might report one of the empirical coefficients determined for a fuel cell in the following manner: "The open-circuit cell potential parameter was determined to be E0,R=0.88 ± 0.6 V, where the uncertainty is the 95% confidence interval on the regression coefficient."
The material below describes methods to determine each of these error estimates, with the exception of confidence interval on the mean from repeated measurements which is discussed here (and is usually included in the output of any descriptive statistics function).