When you start to use Armadillo as a backend for DataSHIELD you can use the DSMolgenisArmadillo
package for research purposes.
There is a default workflow in DataSHIELD to do analysis. These are the steps that you need to take:
First obtain a token from the authentication server to authenticate in DataSHIELD.
# Load the necessary packages. library(dsBaseClient) library(DSMolgenisArmadillo) # specify server url armadillo_url <- "https://armadillo-demo.molgenis.net" # get token from central authentication server token <- armadillo.get_token(armadillo_url) #> [1] "We're opening a browser so you can log in with code 5FW3FV"
Then build a login dataframe and perform the login on the Armadillo server.
The important part is to specify the driver. This should be ArmadilloDriver
.
# build the login dataframe builder <- DSI::newDSLoginBuilder() builder$append( server = "armadillo", url = armadillo_url, token = token, driver = "ArmadilloDriver", profile = "xenon", ) # create loginframe login_data <- builder$build() # login into server conns <- DSI::datashield.login(logins = login_data, assign = FALSE)
You can append multiple servers to the login frame to perform an analysis across multiple cohorts.
To work with DataSHIELD you need to be able to query data. You can do this by assigning data in the Armadillo service.
You can assign values from expressions to symbols.
# assign some data to 'K' datashield.assign.expr(conns = conns, symbol = "K", "c(10,50,100)")
# calculate the mean of 'K' to see that the assignment has worked ds.mean("K", datasources = conns) #> $Mean.by.Study #> EstimatedMean Nmissing Nvalid Ntotal #> armadillo 53.33333 0 3 3 #> #> $Nstudies #> [1] 1 #> #> $ValidityMessage #> ValidityMessage #> armadillo "VALID ANALYSIS"
You can check which tables (data.frame
's) are available on the Armadillo.
datashield.tables(conns) #> $armadillo #> [1] "study1/2_1-core-1_0/nonrep" "study1/2_1-core-1_0/yearlyrep" #> [3] "study1/1_1-outcome-1_0/yearlyrep" "gecko/2_1-core-1_0/trimesterrep" #> [5] "gecko/2_1-core-1_0/nonrep" "gecko/2_1-core-1_0/yearlyrep" #> [7] "gecko/2_1-core-1_0/monthlyrep" "gecko/1_1-outcome-1_0/nonrep" #> [9] "gecko/1_1-outcome-1_0/yearlyrep" "test/data/LT-example-dataset_long-format" #> [11] "test/data/d" "trajectories/data/alspac" #> [13] "trajectories/data/chs" "trajectories/data/bib" #> [15] "trajectories/data/bcg" "trajectories/data/d" #> [17] "trajectories/data/probit" "inma/1_2_urban_ath_1_0/yearly_rep" #> [19] "inma/1_2_urban_ath_1_0/trimester_rep" "inma/1_2_urban_ath_1_0/non_rep" #> [21] "inma/1_1_outcome_ath_1_0/trimester_rep" "inma/1_1_outcome_ath_1_0/non_rep" #> [23] "inma/1_0_outcome_ath_1_0/trimester_rep" "inma/1_0_outcome_ath_1_0/non_rep" #> [25] "longitools/testparquet/LT_example_data" "longitools/mydata/nonrep"
And load data from one of these tables.
# assign table data to a symbol datashield.assign.table( conns = conns, table = "gecko/2_1-core-1_0/nonrep", symbol = "core_nonrep" )
# check the columns in the non-repeated data ds.colnames("core_nonrep", datasources = conns) #> $armadillo #> [1] "row_id" "child_id" "mother_id" "cohort_id" "preg_no" #> [6] "child_no" "coh_country" "recruit_age" "cob_m" "ethn1_m" #> [11] "ethn2_m" "ethn3_m" "agebirth_m_y" "agebirth_m_d" "death_m" #> [16] "death_m_age" "prepreg_weight" "prepreg_weight_mes" "prepreg_weight_ga" "latepreg_weight" #> [21] "latepreg_weight_mes" "latepreg_weight_ga" "preg_gain" "preg_gain_mes" "height_m" #> [26] "height_mes_m" "prepreg_dia" "preg_dia" "preg_thyroid" "preg_fever" #> [31] "preeclam" "preg_ht" "asthma_m" "prepreg_psych" "preg_psych" #> [36] "ppd" "prepreg_smk" "prepreg_cig" "preg_smk" "preg_cig" #> [41] "prepreg_alc" "prepreg_alc_unit" "preg_alc" "preg_alc_unit" "folic_prepreg" #> [46] "folic_preg12" "folic_post12" "parity_m" "preg_plan" "mar" #> [51] "ivf" "outcome" "mode_delivery" "plac_abrup" "cob_p" #> [56] "cob_p_fath" "ethn1_p" "ethn2_p" "ethn3_p" "ethn_p_fath" #> [61] "agebirth_p_y" "agebirth_p_d" "agebirth_p_fath" "death_p" "death_p_age" #> [66] "death_p_fath" "weight_f1" "weight_mes_f1" "weight_f1_fath" "height_f1" #> [71] "height_mes_f1" "height_f1_fath" "dia_bf" "asthma_bf" "psych_bf" #> [76] "smk_p" "smk_cig_p" "smk_fath" "birth_month" "birth_year" #> [81] "apgar" "neo_unit" "sex" "plurality" "ga_lmp" #> [86] "ga_us" "ga_mr" "ga_bj" "birth_weight" "birth_length" #> [91] "birth_head_circum" "weight_who_ga" "plac_weight" "con_anomalies" "major_con_anomalies" #> [96] "cer_palsy" "sibling_pos" "death_child" "death_child_age" "breastfed_excl" #> [101] "breastfed_any" "breastfed_ever" "solid_food" "childcare_intro" "cats_preg" #> [106] "dogs_preg" "cats_quant_preg" "dogs_quant_preg"
You can also specify a table in the login frame and assign the data when you login.
# build the login dataframe builder <- DSI::newDSLoginBuilder() builder$append( server = "armadillo", url = armadillo_url, token = token, driver = "ArmadilloDriver", table = "gecko/2_1-core-1_0/nonrep", profile = "xenon", ) # create loginframe login_data <- builder$build() # login into server conns <- DSI::datashield.login(logins = login_data, assign = TRUE, symbol="core_nonrep")
Before you are working with the data you can subset to a specific range of variables you want to use in the set.
# assign the repeated data to reshape datashield.assign.table( conns = conns, table = "gecko/2_1-core-1_0/yearlyrep", symbol = "core_yearlyrep" ) # check dimensions of repeatead measures ds.dim("core_yearlyrep", datasources = conns) #> $`dimensions of core_yearlyrep in armadillo` #> [1] 19000 34 #> #> $`dimensions of core_yearlyrep in combined studies` #> [1] 19000 34 # subset the data to the first 2 years ds.dataFrameSubset( df.name = "core_yearlyrep", newobj = "core_yearlyrep_1_3", V1.name = "core_yearlyrep$age_years", V2.name = "2", Boolean.operator = "<=" ) #> $is.object.created #> [1] "A data object <core_yearlyrep_1_3> has been created in all specified data sources" #> #> $validity.check #> [1] "<core_yearlyrep_1_3> appears valid in all sources" # check the columns ds.colnames("core_yearlyrep_1_3", datasources = conns) #> $armadillo #> [1] "row_id" "child_id" "age_years" "cohab_" "occup_m_" #> [6] "occupcode_m_" "edu_m_" "occup_f1_" "occup_f1_fath" "occup_f2_" #> [11] "occup_f2_fath" "occupcode_f1_" "occupcode_f1_fath" "occupcode_f2_" "occupcode_f2_fath" #> [16] "edu_f1_" "edu_f1_fath" "edu_f2_" "edu_f2_fath" "childcare_" #> [21] "childcarerel_" "childcareprof_" "childcarecentre_" "smk_exp" "pets_" #> [26] "cats_" "cats_quant_" "dogs_" "dogs_quant_" "mental_exp" #> [31] "hhincome_" "fam_splitup" "famsize_child" "famsize_adult" # check dimensions again ds.dim("core_yearlyrep_1_3", datasources = conns) #> $`dimensions of core_yearlyrep_1_3 in armadillo` #> [1] 3000 34 #> #> $`dimensions of core_yearlyrep_1_3 in combined studies` #> [1] 3000 34
# strip the redundant columns ds.dataFrame( x = c("core_yearlyrep_1_3$child_id", "core_yearlyrep_1_3$age_years", "core_yearlyrep_1_3$dogs_", "core_yearlyrep_1_3$cats_", "core_yearlyrep_1_3$pets_"), completeCases = TRUE, newobj = "core_yearlyrep_1_3_stripped", datasources = conns ) #> $is.object.created #> [1] "A data object <core_yearlyrep_1_3_stripped> has been created in all specified data sources" #> #> $validity.check #> [1] "<core_yearlyrep_1_3_stripped> appears valid in all sources"
In general you need 2 methods to work with data that is stored in long format,
the reshape
and merge
functions in DataSHIELD. You can reshape data with
the Armadillo to transform data from wide-format to long-format and
vice versa.
You can do this using the ds.reshape
function:
# reshape the data for the wide-format variables (yearlyrep) ds.reShape( data.name = "core_yearlyrep_1_3_stripped", timevar.name = "age_years", idvar.name = "child_id", v.names = c("pets_", "cats_", "dogs_"), direction = "wide", newobj = "core_yearlyrep_1_3_wide", datasources = conns ) #> $is.object.created #> [1] "A data object <core_yearlyrep_1_3_wide> has been created in all specified data sources" #> #> $validity.check #> [1] "<core_yearlyrep_1_3_wide> appears valid in all sources"
# show the reshaped columns of the new data frame ds.colnames("core_yearlyrep_1_3_wide", datasources = conns) #> $armadillo #> [1] "child_id" "pets_.0" "cats_.0" "dogs_.0" "pets_.1" "cats_.1" "dogs_.1" "pets_.2" "cats_.2" "dogs_.2"
When you reshaped and subsetted the data you often need to merge your dataframe
with others to get your analysis dataframe. You can do this using the ds.merge
function:
# merge non-repeated table with wide-format repeated table # make sure the disclosure measure regarding stringshort is set to '100' ds.merge( x.name = "core_nonrep", y.name = "core_yearlyrep_1_3_wide", by.x.names = "child_id", by.y.names = "child_id", newobj = "analysis_df", datasources = conns ) #> $is.object.created #> [1] "A data object <analysis_df> has been created in all specified data sources" #> #> $validity.check #> [1] "<analysis_df> appears valid in all sources"
ds.colnames("analysis_df", datasources = conns) #> $armadillo #> [1] "child_id" "row_id" "mother_id" "cohort_id" "preg_no" #> [6] "child_no" "coh_country" "recruit_age" "cob_m" "ethn1_m" #> [11] "ethn2_m" "ethn3_m" "agebirth_m_y" "agebirth_m_d" "death_m" #> [16] "death_m_age" "prepreg_weight" "prepreg_weight_mes" "prepreg_weight_ga" "latepreg_weight" #> [21] "latepreg_weight_mes" "latepreg_weight_ga" "preg_gain" "preg_gain_mes" "height_m" #> [26] "height_mes_m" "prepreg_dia" "preg_dia" "preg_thyroid" "preg_fever" #> [31] "preeclam" "preg_ht" "asthma_m" "prepreg_psych" "preg_psych" #> [36] "ppd" "prepreg_smk" "prepreg_cig" "preg_smk" "preg_cig" #> [41] "prepreg_alc" "prepreg_alc_unit" "preg_alc" "preg_alc_unit" "folic_prepreg" #> [46] "folic_preg12" "folic_post12" "parity_m" "preg_plan" "mar" #> [51] "ivf" "outcome" "mode_delivery" "plac_abrup" "cob_p" #> [56] "cob_p_fath" "ethn1_p" "ethn2_p" "ethn3_p" "ethn_p_fath" #> [61] "agebirth_p_y" "agebirth_p_d" "agebirth_p_fath" "death_p" "death_p_age" #> [66] "death_p_fath" "weight_f1" "weight_mes_f1" "weight_f1_fath" "height_f1" #> [71] "height_mes_f1" "height_f1_fath" "dia_bf" "asthma_bf" "psych_bf" #> [76] "smk_p" "smk_cig_p" "smk_fath" "birth_month" "birth_year" #> [81] "apgar" "neo_unit" "sex" "plurality" "ga_lmp" #> [86] "ga_us" "ga_mr" "ga_bj" "birth_weight" "birth_length" #> [91] "birth_head_circum" "weight_who_ga" "plac_weight" "con_anomalies" "major_con_anomalies" #> [96] "cer_palsy" "sibling_pos" "death_child" "death_child_age" "breastfed_excl" #> [101] "breastfed_any" "breastfed_ever" "solid_food" "childcare_intro" "cats_preg" #> [106] "dogs_preg" "cats_quant_preg" "dogs_quant_preg" "pets_.0" "cats_.0" #> [111] "dogs_.0" "pets_.1" "cats_.1" "dogs_.1" "pets_.2" #> [116] "cats_.2" "dogs_.2"
When you finished building your analysis frame you can save it using workspaces.
There are a variety of analysis you can perform in DataSHIELD. You can perform basic methods such as summary statistics and more advanced methods such as GLM.
You execute a summary on the a variable within you analysis frame. It will return summary statistics.
ds.summary("analysis_df$pets_.1", datasources = conns) #> $armadillo #> $armadillo$class #> [1] "numeric" #> #> $armadillo$length #> [1] 1000 #> #> $armadillo$`quantiles & mean` #> 5% 10% 25% 50% 75% 90% 95% Mean #> 8.000 15.000 32.750 61.000 90.000 108.000 113.000 60.954
When you finished the analysis dataframe, you can perform the actual analysis.
You can use a wide variety of functions. The example below is showing the glm
.
datashield.assign.table( conns = conns, table = "gecko/1_1-outcome-1_0/nonrep", symbol = "outcome_nonrep" ) armadillo_glm <- ds.glm( formula = "asthma_ever_CHICOS~pets_preg", data = "outcome_nonrep", family = "binomial", datasources = conns )
Do the meta analysis and install prerequisites.
if (!require('metafor')) install.packages('metafor')
yi <- c(armadillo_glm$coefficients["pets_preg", "Estimate"]) sei <- c(armadillo_glm$coefficients["pets_preg", "Std. Error"]) res <- metafor::rma(yi, sei = sei) res #> #> Random-Effects Model (k = 1; tau^2 estimator: REML) #> #> tau^2 (estimated amount of total heterogeneity): 0 #> tau (square root of estimated tau^2 value): 0 #> I^2 (total heterogeneity / total variability): 0.00% #> H^2 (total variability / sampling variability): 1.00 #> #> Test for Heterogeneity: #> Q(df = 0) = 0.0000, p-val = 1.0000 #> #> Model Results: #> #> estimate se zval pval ci.lb ci.ub #> -0.1310 0.1267 -1.0343 0.3010 -0.3793 0.1173 #> #> --- #> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 metafor::forest(res, xlab = "OR", transf = exp, refline = 1, slab = c("armadillo_glm"))
You can directly create figures with the DataSHIELD methods. For example:
# create histogram ds.histogram(x = "core_nonrep$coh_country", datasources = conns)
#> $breaks #> [1] 35.31138 116.38319 197.45500 278.52680 359.59861 440.67042 521.74222 602.81403 683.88584 764.95764 846.02945 #> #> $counts #> [1] 106 101 92 103 106 104 105 101 113 69 #> #> $density #> [1] 0.0013074829 0.0012458092 0.0011347965 0.0012704787 0.0013074829 0.0012828134 0.0012951481 0.0012458092 0.0013938261 #> [10] 0.0008510974 #> #> $mids #> [1] 75.84729 156.91909 237.99090 319.06271 400.13451 481.20632 562.27813 643.34993 724.42174 805.49355 #> #> $xname #> [1] "xvect" #> #> $equidist #> [1] TRUE #> #> attr(,"class") #> [1] "histogram"
# create a heatmap ds.heatmapPlot(x = "analysis_df$pets_.1", y = "analysis_df$dogs_.1", datasources = conns)
# logout datashield.logout(conns)
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.