Unfortunately, errors sometimes slip through despite my best efforts to catch them before manuscript submission and publication. Since there's no good way to correct minor errors in published work at present, I'm disclosing them here.
If you spot an error in any of my papers, I'd be grateful if you could let me know.
If you spot an error in any of my papers, I'd be grateful if you could let me know.
1) Incorrect parameter values listed in supplementary materials
Paper: MK Thomas, M Aranguren-Gassis, CT Kremer, MR Gould, K Anderson, CA Klausmeier & E Litchman. (2017). Temperature-nutrient interactions
exacerbate sensitivity to warming in phytoplankton. Global Change Biology. 23, 3269-3280. [pdf]
Two parameter values reported in a supplementary table were off by orders of magnitude (they belong to a reference model that was not the focus of our study). This reference model is a challenge to fit to data because of its complex structure. To ease the fitting procedure and obtain more accurate results, we had rescaled some values. We forgot to return the values to the original scale when entering them in the supplementary table.
The precise nature of the error (in Table S1) is:
Paper: MK Thomas, M Aranguren-Gassis, CT Kremer, MR Gould, K Anderson, CA Klausmeier & E Litchman. (2017). Temperature-nutrient interactions
exacerbate sensitivity to warming in phytoplankton. Global Change Biology. 23, 3269-3280. [pdf]
Two parameter values reported in a supplementary table were off by orders of magnitude (they belong to a reference model that was not the focus of our study). This reference model is a challenge to fit to data because of its complex structure. To ease the fitting procedure and obtain more accurate results, we had rescaled some values. We forgot to return the values to the original scale when entering them in the supplementary table.
The precise nature of the error (in Table S1) is:
the following values:
A = 28.75
B = 0.518
should have been entered as:
A = 28.75 x 102
B = 0.518 x 10-4
A = 28.75
B = 0.518
should have been entered as:
A = 28.75 x 102
B = 0.518 x 10-4
None of our results or findings are affected in any way as the analysis was done appropriately; the error is in the values we reported only.
I contacted the journal about this but my request to have a correction published was denied because the error did not affect the results of the paper.
2. Improper evaluation of predictability/forecast accuracy
Paper: MK Thomas, S Fontana, M Reyes, M Kehoe & F Pomati (2018). The predictability of a lake phytoplankton community, over time-scales of hours to
years. Ecology Letters. 21, 619-628. [pdf]
In this study, I do not believe that I accurately quantified predictability/forecast ability in Figure 3. I thought out-of-bag accuracy would be a reasonable way to quantify this, but my approach suffers from what in the machine learning world is called 'data leakage'. What I should have done was perform rolling window forecasts and evaluate performance on those, but I was unaware of this at the time.
The qualitative patterns and other results are likely robust but I think I overestimated predictability.
Paper: MK Thomas, S Fontana, M Reyes, M Kehoe & F Pomati (2018). The predictability of a lake phytoplankton community, over time-scales of hours to
years. Ecology Letters. 21, 619-628. [pdf]
In this study, I do not believe that I accurately quantified predictability/forecast ability in Figure 3. I thought out-of-bag accuracy would be a reasonable way to quantify this, but my approach suffers from what in the machine learning world is called 'data leakage'. What I should have done was perform rolling window forecasts and evaluate performance on those, but I was unaware of this at the time.
The qualitative patterns and other results are likely robust but I think I overestimated predictability.
