Frequently Asked Questions

Why use transform = 'x+1' with prepareData2Fit()

Using transform = 'x+1' with prepareData2Fit() applies a simple transformation by adding 1 to all count values. This is particularly useful because zeros in the data can sometimes pose challenges for analyses. Applying this transformation can rescue the fitting process for some genes without negatively affecting the results. Additionally, this transformation is sometimes indispensable for data normalization, helping to avoid the error: “Every gene contains at least one zero, cannot compute log geometric means.”

Why use row_threshold = 10 with prepareData2Fit()

Setting row_threshold = 10 in prepareData2Fit() filters out genes with very low expression levels, aiming to enhance the quality of the data used for fitting models. This thresholding approach removes genes with sparse expression profiles, which may not provide sufficient information for reliable model estimation, leading to more robust and accurate analyses. Adjust the threshold as needed based on the characteristics of your data.

Why mu has no effect within simulation

In HTRfit, the parameter mu in init_variable() or add_interaction() represents the mean of the distribution used to simulate data. If mu seems to have no effect within your simulation, it could be due to the scaling of counts, performed to control sequencing depth. Indeed, the control of the sequencing depth is overriding the influence of mu on the (Intercept) values. That’s why we recommend to keep mu = 0.