Discrepancy in calculation of cumulative migration probability

[ariloytynoja]

Hello,

I'm confused about the correct way of computing M(t).

In MSMC_IM.py, they are computed using the right boundaries and then outputted in the file *.MSMC_IM.estimates.txt:

MSMC-IM/MSMC_IM.py

Line 79 in e6c5d3b

CumulativeDF = MSMC_IM_funcs.cumulative_Symmigproportion(right_boundaries, m_List)

However, in plot_utils.py, they are computed using the left boundaries as only those are included in *.MSMC_IM.estimates.txt:

time_boundaries,IM_N1s,IM_N2s,m_t_s = read_from_MSMC_IM(args.Input)
if args.samegrid:
    t = time_boundaries
    CDF_t = cumulative_Symmigproportion(time_boundaries, m_t_s)

MSMC-IM/plot_utils.py

Line 86 in e6c5d3b

CDF_t = cumulative_Symmigproportion(time_boundaries, m_t_s)

As the first left boundary is zero, a consequence of the latter is that integ = 2 * m[0] * 0 is always zero and the first time period is ignored:

MSMC-IM/plot_utils.py

Line 44 in e6c5d3b

integ = 2 * m[0] * time_boundaries[0]

Confusingly, the plotting function in MSMC_IM.py is different from that in plot_utils.py, and erroneous at least if xlog is used and the first x-value (log(0)) is not defined:

MSMC-IM/MSMC_IM.py

Line 190 in e6c5d3b

plot(left_boundaries, CumulativeDF, label='Infer: CDF', c='orange')

There are now three ways of plotting M(t):

• in plot_utils.py for making plots from the *.estimates.txt data but recomputing the M values
• in MSMC_IM.py for plots in *.fittingdetails.xlog.pdf
• user-made plots using the M values in *.estimates.txt

I personally would plot M(t) with the column 'M' from *.MSMC_IM.estimates.txt and the right boundaries. One could add zeros in the vectors to start it from the corner. However, this differs from the plots generated by MSMC-IM.

Can you please tell your opinion on which one of these is the right way to do the plotting?

Best regards,

Ari

[wangke16]

Hi, Regarding time boundaries, I think you should check —samegrid function in plot_utils.py from MSMC-IM repository.

…

On 17 Apr 2024, at 11:21, Ari Löytynoja ***@***.***> wrote: Hello, I'm confused about the correct way of computing M(t). In MSMC_IM.py, they are computed using the right boundaries and then outputted in the file *.MSMC_IM.estimates.txt: https://github.com/wangke16/MSMC-IM/blob/e6c5d3b13d10131a6aec7b5a9fd792a3ce09bec9/MSMC_IM.py#L79 <https://github.com/wangke16/MSMC-IM/blob/e6c5d3b13d10131a6aec7b5a9fd792a3ce09bec9/MSMC_IM.py#L79> However, in plot_utils.py, they are computed using the left boundaries as only those are included in *.MSMC_IM.estimates.txt: time_boundaries,IM_N1s,IM_N2s,m_t_s = read_from_MSMC_IM(args.Input) if args.samegrid: t = time_boundaries CDF_t = cumulative_Symmigproportion(time_boundaries, m_t_s) https://github.com/wangke16/MSMC-IM/blob/e6c5d3b13d10131a6aec7b5a9fd792a3ce09bec9/plot_utils.py#L86 <https://github.com/wangke16/MSMC-IM/blob/e6c5d3b13d10131a6aec7b5a9fd792a3ce09bec9/plot_utils.py#L86> As the first left boundary is zero, a consequence of the latter is that integ = 2 * m[0] * 0 is always zero and the first time period is ignored: https://github.com/wangke16/MSMC-IM/blob/e6c5d3b13d10131a6aec7b5a9fd792a3ce09bec9/plot_utils.py#L44 <https://github.com/wangke16/MSMC-IM/blob/e6c5d3b13d10131a6aec7b5a9fd792a3ce09bec9/plot_utils.py#L44> Confusingly, the plotting function in MSMC_IM.py is different from that in plot_utils.py, and erroneous at least if xlog is used and the first x-value (log(0)) is not defined: https://github.com/wangke16/MSMC-IM/blob/e6c5d3b13d10131a6aec7b5a9fd792a3ce09bec9/MSMC_IM.py#L190 <https://github.com/wangke16/MSMC-IM/blob/e6c5d3b13d10131a6aec7b5a9fd792a3ce09bec9/MSMC_IM.py#L190> There are now three ways of plotting M(t): in plot_utils.py for making plots from the *.estimates.txt data but recomputing the M values in MSMC_IM.py for plots in *.fittingdetails.xlog.pdf user-made plots using the M values in *.estimates.txt I personally would plot M(t) with the column 'M' from *.MSMC_IM.estimates.txt and the right boundaries. One could add zeros in the vectors to start it from the corner. However, this differs from the plots generated by MSMC-IM. Can you please tell your opinion on which one of these is the right way to do the plotting? Best regards, Ari — Reply to this email directly, view it on GitHub <#13>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGLYZY62PWKK2EFPPY3YKTDY5Y5IPAVCNFSM6AAAAABGK5KP26VHI2DSMVQWIX3LMV43ASLTON2WKOZSGI2DOOBQG42TEMI>. You are receiving this because you are subscribed to this thread.